# State of AI 2026 Final SEG
## Main Cover
State of AI
Latin America
2026
4th edition
## Letter
A Letter from HI
**Three** years ago, the central question around artificial intelligence in Latin America was whether the region would adopt it.
Today, that question feels largely settled. AI has moved from experimentation to everyday use. Founders are building with it from day one, employees increasingly rely on it to do their jobs, and corporations across the region are moving beyond pilots into real deployment.
The story now is agents.
For the first time, software can do more than answer questions or generate content. AI agents can pursue objectives, navigate workflows, use tools, and increasingly operate with limited human supervision. This shift has the potential to reshape not only products, but also how companies, institutions, and entire industries are organized.
Latin America enters this transition from a position of strength. The region's AI adoption rates rival those of far wealthier economies, while its entrepreneurs have decades of experience building in complex, resource-constrained environments. We believe that combination may prove to be an advantage as the next generation of AI-native companies emerges.
This fourth edition of the State of AI in Latin America explores that transition. Through data, interviews, and conversations with founders, operators, researchers, and investors, we examine how AI is reshaping organizations across the region and where the most important opportunities may emerge next.
Jimena Pardo
Ariel Arrieta
## Acknowledgements
Credits & Collaborators
Acknowledgements
Presented by
https://elevenlabs.io/agents?utm_source=stateofai&utm_medium=referral&utm_campaign=state_of_ai_latam_2026&utm_term=b2b
With the support of
| Logo | Alt |
| --- | --- |
|  | tess |
|  | faces |
|  | LAVCA |
Prepared by our editorial team & AI collaborators
ChatGPT · Claude · Gemini - Tess - Faces
| Name | Role | Photo |
| --- | --- | --- |
| Antonia Bezanilla | Research |  |
| Alejandro Cercio | Research |  |
| Leonardo Coz | Principal |  |
| Jimena Pardo | Managing Partner |  |
| Federico Antoni | Managing Partner |  |
© Hi Ventures 2026 / hi.vc June 2026
## Insights List
Key insights
| Index | Heading | Description |
| --- | --- | --- |
| 01 | **Latin America evolved from experimentation to infrastructure.** | What was a pilot in 2024 is now production-grade with agentic frameworks at a speed of adoption that took cloud and mobile twice as long. |
| 02 | Agentic big bang. | 53% of startups already deploy agents in production and 30% of corporates do too, but human-in-the-loop remains the dominant autonomy model across both segments. |
| 03 | AI Readiness is converging across the region. | Chile (72.5) and Argentina (71.0) now lead the Hi AI Readiness Index, while Brazil jumped from 62.5 to 70.2 — the largest year-over-year gain. |
| 04 | Talent, not capital, is the binding constraint. | Lack of technical talent is a top-3 barrier for both startups (23%) and corporates (28%), even as 51% of startups allocate over 10% of their operating budget to AI. |
| 05 | The VC ecosystem is going all-in: 91% of VCs use AI to analyze deals. | (vs. 45% two years ago) and 61% say over 60% of their new investments have AI embedded in the core product. |
## Spotlight
About Hi Ventures
HI Ventures is an early-stage venture capital firm backing Latin American founders building innovative global companies. We invest with deep conviction in technology, long-term founder partnerships, and disciplined capital allocation.
Federico Antoni and Jimena Pardo, co-founders of Hi Ventures, have built together for over a decade, from Hi's early investment in Jimena's startup to leading the firm today. Jimena previously led product growth initiatives at Meta and is both a Kauffman Fellow and Endeavor Entrepreneur. Federico teaches entrepreneurship and AI at Stanford Graduate School of Business and has backed category-defining companies built from Latin America to the world.
Since 2023, Hi Ventures has focused exclusively on AI. We invest in founders building AI-native companies, but our work extends beyond investing. We believe artificial intelligence will reshape how startups are built, how corporations operate, and how institutions deliver services across Latin America. Understanding that transformation requires more than capital. It requires research, dialogue, and continuous engagement with the people driving adoption.
The State of AI in Latin America is part of that effort. Through annual surveys, interviews, company analysis, and ongoing collaboration with founders, executives, investors, and policymakers, we seek to build a clearer picture of how AI is spreading across the region, where value is being created, and what challenges remain ahead.
Hi Ventures has invested in more than 50 startups, with realized outcomes including Cornershop (Uber), R2 (Ant Group), and Atlas (Remote), and long-term positions in companies such as Fintual, Mendel, and Nuvocargo.
## Contents
Table of Contents
Inside the Report
| No. | Section | Description | Color | Face |
| --- | --- | --- | --- | --- |
| 01 | AI year in review | The agentic big bang | #B1BAFF | talent |
| 02 | Hi on AI | our thesis for the AI era | #FD88C0 | landscape |
| 03 | Corporates | from copilots to workflows | #58BED6 | capital |
| 04 | Startups | built for the agent era | #50C8BA | infrastructure |
| 05 | Venture Capital | investing in the age of agents | #ABC9CE | outlook |
## Download Report
Get the full report
Download the complete State of AI in Latin America 2026 study — data, interviews, and insights from founders, operators, and investors across the region.
Download PDF
https://d6yvfl55smr7u.cloudfront.net/assets/36o5bnh9-1782398749368-state-of-ai-2026.pdf
PDF · 4th edition · June 2026
:icon-arrow-down-to-line:
## Blank slide
Talk to State of AI
Connecting…
End call
Listening…
Speaking…
Microphone access needed to start the call
https://elevenlabs.io/app/talk-to?agent_id=agent_9601kvy6asx5e6kaw9zjb3xj9b9x&branch_id=agtbrch_9501kvy6atehf27rmvy6j5tdad37
https://elevenlabs.io/agents?utm_source=stateofai&utm_medium=referral&utm_campaign=state_of_ai_latam_2026&utm_term=b2b
## Interlude
"Nobody knows anything..."
William Goldman
American novelist
## Landscape
AI year in review
01
## AI Revolution Speed
AI is emerging faster than previous technological revolutions
Previous technological revolutions such as railroads, electrification, and the internet reshaped economies over decades. In contrast, nearly** US$1 trillion has already been committed to AI infrastructure and capabilities, **suggesting that this technological wave may unfold much faster and potentially allow countries to accelerate the development of productive capabilities.
Cumulative capital investment by infrastructure programs (inflation-adjusted)
2026 (planned)
Years from start of program
Sources: Company reports, Epoch AI, FHWA, NASA, CRS, GAO, Brookings
| Name | Caption | Points (year,value;...) | Label Year | Label Value | Highlight |
| --- | --- | --- | --- | --- | --- |
| AI Capital Investments | ≈$930B in 6 years | 0,0;1,45;2,130;3,250;4,500;5,700;6,930 | 7 | 960 | Yes |
| Marshall Plan | $170B, 4yr | 0,0;1,55;2,100;3,140;4,170 | 2 | 215 | No |
| Apollo Program | $257B, 14yr | 0,0;4,60;8,140;11,210;14,257 | 9 | 305 | No |
| F-35 Program | $400B, 25yr | 0,0;6,90;12,190;18,290;25,400 | 17 | 445 | No |
| Interstate Highway System | $620B, 37yr | 0,0;8,90;16,210;25,390;31,490;37,620 | 28 | 665 | No |
| Int. Space Station | $150B, 27yr | 0,0;7,30;14,70;21,110;27,150 | 29 | 200 | No |
| US Railroads | $550B, 71yr | 0,0;15,90;30,200;40,250;50,340;60,440;71,550 | 60 | 600 | No |
## AI Capability Acceleration
AI models are becoming dramatically more capable
Frontier AI capability is accelerating. The best models now improve at +15.5 points per year on standard benchmarks, nearly double the +8.3 pace set between 2022 and early 2024.
Source: Goldman Sachs, AI Agents to Boost Productivity and Size of Software Market (Jul 2025). Data: Gartner (Oct 2024)
10 © Hi Ventures 2026 / hi.vc June 2026
+8.3 points per year
+15.5 points per year
| Year | Value | Label | Type (frontier/standard/historic) |
| --- | --- | --- | --- |
| 2022.0 | 104 | | standard |
| 2022.2 | 110 | | standard |
| 2022.3 | 115 | | standard |
| 2023.1 | 125 | | standard |
| 2024.1 | 131 | | standard |
| 2024.9 | 144 | o1 (high) | frontier |
| 2025.1 | 139 | DeepSeek R-1 | frontier |
| 2025.4 | 152 | GPT-5 Pro | frontier |
| 2025.6 | 148 | Grok 4 | frontier |
| 2025.8 | 154 | Gemini 3 Flash | frontier |
| 2025.9 | 151 | Claude Opus 4.5 | frontier |
| 2024.5 | 118 | | historic |
| 2024.7 | 122 | | historic |
| 2024.2 | 128 | | historic |
| 2025.2 | 134 | | historic |
| 2025.4 | 137 | | historic |
| 2023.8 | 112 | | historic |
## Acceleration Table
Four years of steep acceleration: what's changed since our first report
2023
2024
2025
2026
% Change
(2023-26)
| Metric | Unit | 2023 | 2024 | 2025 | 2026 | Direction | Change |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Cost per 1M tokens, | output tokens in USD | 60 | 30 | 10 | ~3 | down | 95% |
| Context window, | max tokens per prompt | 32k | 128k | 1M | 12M | up | 375x |
| Inference latency, | ms | 1,300 | 500 | 100 | ~100 | down | 92% |
| NVIDIA H100, | unit price in USD | ~40,000 | ~30,000 | ~25,000 | ~15,000 | down | 63% |
| Tools per API call, | max parallel | 1 | ~10 | ~50 | 128+ | up | 128x |
| Persistent memory, | tokens across sessions | 0 | ~2k | ~20k | ~200k | up | 200k |
Sources: OpenAI, Anthropic, Google, NVIDIA, BFCL (Berkeley). Cost based on output pricing (GPT-4 class)
© Hi Ventures 2026 / hi.vc June 2026
8
## Year in Review Timeline
The last twelve months redefined the AI landscape
2025
2026
Drag to explore the timeline
| Date | Heading | Description | Logo |
| --- | --- | --- | --- |
| May 22nd | Anthropic releases Claude 4 (Opus 4 & Sonnet 4), | setting new coding benchmarks and launching Claude Code for agentic development |  |
| July 9th | Nvidia becomes first $4T company, | driven by insatiable demand for AI chips and data center infrastructure |  |
| August 7th | OpenAI launches GPT-5, | the year's defining model release, unifying reasoning and general-purpose capabilities in one system |  |
| November 18th | Google launches Gemini 3, | part of an unprecedented 25-day sprint where four frontier models launched in rapid succession |  |
| November 24th | OpenClaw launches, | an open-source AI agent by Peter Steinberger that goes viral with 247K+ GitHub stars, sparking a global agentic AI movement |  |
| December 11th | White House AI executive order | establishes a national AI policy framework with unified standards, preempting state-level regulations |  |
| March 25th | MCP hits 97M installs, | going from experimental protocol to universal AI standard in 16 months. Every major provider now ships MCP tooling |  |
| March 31st | OpenAI raises $122B at $852B valuation, | the largest AI funding round in history, led by Amazon ($50B), Nvidia ($30B), and SoftBank ($30B) |  |
| April 7th | Anthropic previews Claude Mythos, | a new class of AI so powerful in cybersecurity it required a controlled release via Project Glasswing |  |
| May 15th | ChatGPT crosses 1B weekly users, | becoming the fastest app in history to reach that milestone, with 2.5B daily prompts and 50M paying subscribers |  |
| June 1st | Anthropic and OpenAI file S-1s | within a week of each other, setting up the first trillion-dollar AI IPO race with combined valuations exceeding $1.8T |  |
Sources: OpenAI, Anthropic, Google, NVIDIA, TechCrunch, Financial Times, Reuters, Business Insider, GitHub
9
## Agent Software Category
Agents are emerging as a new software category
Forecast of software total addressable market (TAM) SaaS vs. AI agents
Agentic big bang
SaaS TAM ($ bn)
Agent TAM ($ bn)
OpenClaw became** the fastest-growing open-source project in history, **signaling unprecedented developer interest in autonomous agents.
Claude Code demonstrated that** AI could move beyond content generation** and reliably execute complex engineering workflows.
Together, these advances marked what we call the Agentic Big Bang. **Agents moved from experimental tools to practical coworkers,** expanding the range of tasks software can perform and potentially creating a new software category.
Source: Goldman Sachs, AI Agents to Boost Productivity and Size of Software Market (Jul 2025). Data: Gartner (Oct 2024)
| Year | SaaS TAM | Total TAM |
| --- | --- | --- |
| 2025 | 30 | 30 |
| 2026 | 28 | 32 |
| 2027 | 26 | 37 |
| 2028 | 23 | 43 |
| 2029 | 20 | 47 |
| 2030 | 18 | 50 |
## New Capabilities
New capabilities matter more than automation
The opportunity is not simply to automate existing tasks**.** As Erik Brynjolfsson argues, the most important **impact of AI may come from entirely new capabilities and workflows that did not exist before.**
The first generation of AI helped humans generate answers. **The emerging generation of agents can execute tasks**, evaluate results, and iteratively improve their performance.
As a result, the most valuable opportunities may lie in workflows where success can be measured, verified, and continuously optimized.
14 © Hi Ventures 2026 / hi.vc June 2026
Sources: Erik Brynjolfsson, "The Turing Trap" (2022), Stanford HAI; Boris Cherny, Anthropic (2025); Prompt Era / Loop Era framework
## Expert Insight
AGENTIC DISCOVERY
**Agentic discovery:** Insights from Sofía Elizondo
“
3 in 4 health-innovation efforts fail today. Our job is to make nature's chemistry searchable so more of them succeed.
Sofía Elizondo
CHAIRMAN, BRIGHTSEED
Source: Hi Ventures interview with Sofía Elizondo
| Heading | Body |
| --- | --- |
| Using AI to rediscover nature's pharmacy | Humans carry many receptors that are activated by plants and fungi, so the molecules they produce can act directly on human biology, which is why most small-molecule medicines originate in nature. Yet science had only ever catalogued around 400,000 natural molecules, a tiny fraction of what exists. Brightseed’s Forager platform uses AI to predict a molecule’s structure, the sources that carry it and the receptors it activates, mapping 21 million molecules to date.
|
| Hummingbird: agentic AI for scientific discovery | Brightseed built Hummingbird on top of that data: an agentic platform with dedicated agents for discovery, formulation, IP, and regulatory work that customers directly engage with. The stakes are high, because today roughly 75% of life-science innovation fails. Brightseed has proven it can double success rates in half the time, and already has products in market, including bioactives from for blood-sugar and digestive health.
|
| Why Latin America can leapfrog | With radical technological change new models can emerge. Latin America holds a rare mix of traditional medicinal knowledge, some of the highest biodiversity on earth (Mexico, Brazil, Peru), and urgent health needs. With these AI tools available, the window is now to build a homegrown health advantage.
|
## Productivity Gap
The productivity gap is part of our investment thesis
Total factor productivity, indexed (2012=100): US vs. select Latin American economies
Since 2012, total factor productivity has risen by roughly 6% in the United States but fallen across much of Latin America, ranging from -9% in Mexico to -18% in Argentina.** This growing productivity gap remains one of the region’s largest barriers to economic growth.**
AI has the potential to change the equation. By augmenting workers, automating workflows, and enabling new organizational models,** it could help Latin America achieve productivity gains **that have proven elusive for decades.
16 © Hi Ventures 2026 / hi.vc June 2026
Sources: Penn World Table 11.0 via Our World in Data (2012=100 normalization), IMF WP/24/219 (2024), WEF/McKinsey "Latin America in the Intelligent Age" (2026)
| Name | Color token (CSS var) | End label | Yearly values (comma separated, 2012-2023) |
| --- | --- | --- | --- |
| US | --color-4 | +6% | 100,100,100.5,101,100.8,101.5,102,103,103.5,106.3,104.5,105.5 |
| Brazil | --color-5 | -17% | 100,96,92,89,87.7,88,87,86,88.5,85,82,83 |
| Mexico | --color-1 | -9% | 100,98,99,98,96,94.5,93,92,91,90.5,91,91 |
| Chile | --color-3 | -11% | 100,99,95,94,93.5,94.5,93,92,95.8,94,90,89 |
| Argentina | --color-2 | -18% | 100,98,95,94.5,93.5,92,90,86,84.5,85,88,82 |
## Blank slide
The AI paradox: automation increases demand for talent
Indeed job postings index: software development and overall market (2024-2026)
Software Development
All Postings
Software development postings rebounded sharply from early 2025, diverging from a flat overall job market. The signal is that **companies aren't hiring broadly, they're hiring specifically for AI-capable roles.**
Sources: Bloomberg, Citadel Securities, Indeed Job Postings For Software Engineers
© Hi Ventures 2026 / hi.vc June 2026
| Month | All Postings | Software Development |
| --- | --- | --- |
| Jan-24 | 73 | 119 |
| Feb-24 | 72.5 | 117 |
| Mar-24 | 72 | 116.5 |
| Apr-24 | 71 | 116 |
| May-24 | 70 | 114 |
| Jun-24 | 69.3 | 113 |
| Jul-24 | 68.3 | 113 |
| Aug-24 | 67 | 111.5 |
| Sep-24 | 66.5 | 110.5 |
| Oct-24 | 66.5 | 111 |
| Nov-24 | 66 | 109 |
| Dec-24 | 65 | 107 |
| Jan-25 | 64 | 105 |
| Feb-25 | 63.5 | 103 |
| Mar-25 | 63.5 | 101.5 |
| Apr-25 | 63 | 100 |
| May-25 | 63 | 101 |
| Jun-25 | 63.2 | 103 |
| Jul-25 | 63.5 | 105 |
| Aug-25 | 63.8 | 108 |
| Sep-25 | 64 | 110 |
| Oct-25 | 64.3 | 113 |
| Nov-25 | 64.2 | 115 |
| Dec-25 | 63.8 | 116.5 |
| Jan-26 | 64 | 118 |
## Data Spectrum
HI AI READINESS INDEX
A tight race and converging maturity across the region
The Hi AI Readiness Index captures **how prepared** a country's organizations are **to adopt, scale, and govern AI**.
A composite score from 0 to 100 across five pillars: **Strategy** (30%), **Investment** (20%), **Talent & Culture** (20%), **Infrastructure & Data** (15%), and **Governance** (15%).
**Year-over-year, the region is maturing fast.** Chile (72.5) and Argentina (71.0) now lead the index.
| Pillar |
| --- |
| Strategy |
| Investment |
| Talent & Culture |
| Infra & Data |
| Governance |
| Rank |
| --- |
| 1 |
| 2 |
| 3 |
| 4 |
| 5 |
| Country | Strategy | Investment | Talent & Culture | Infra & Data | Governance | Overall score | Overall rank |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Chile | 1 | 3 | 4 | 3 | 1 | 72.5 | 1 |
| Argentina | 1 | 5 | 1 | 2 | 1 | 71.0 | 2 |
| Brazil | 1 | 2 | 3 | 4 | 1 | 70.2 | 3 |
| Mexico | 5 | 3 | 2 | 5 | 1 | 67.5 | 4 |
| Colombia | 1 | 1 | 5 | 1 | 5 | 53.1 | 5 |
Medians used to reduce sample-size and outlier bias. Results reflect the survey sample, not the full ecosystem.
## Robotics Manufacturing
Robotics could upgrade Latin America’s manufacturing base
**Physical AI is moving beyond isolated robots into entire industrial systems,** including manufacturing, logistics, maintenance, quality control, mining, and agriculture.
These are sectors where Latin America already has global relevance, from manufacturing in Mexico and Brazil to mining in Chile and Peru, and agriculture in Argentina and Brazil.
The opportunity is not only greater output, **but more intelligent, flexible, and higher-productivity **operations across the region’s core industries.
Direct manufacturing
Indirect manufacturing
| Number | Label | Group |
| --- | --- | --- |
| 1 | Adaptive welding for high-mix volume parts | MATERIALS |
| 2 | Surface treatment with AI vision system for real-time adjustment | MATERIALS |
| 3 | Versatile assembly with parts variation | ASSEMBLY |
| 4 | Dexterous cable assembly | ASSEMBLY |
| 5 | Force-sensitive robotic arms for precise component insertion | ASSEMBLY |
| 6 | Automated packaging with adaptable package sizes | PACKAGING |
| 7 | Automated boxing of parts with AI vision for optimal part arrangement | PACKAGING |
| Number | Label | Group |
| --- | --- | --- |
| 8 | Raw material load/unload | WAREHOUSE & LOGISTICS |
| 9 | Tote load/unload | WAREHOUSE & LOGISTICS |
| 10 | Autonomous shelving | WAREHOUSE & LOGISTICS |
| 11 | Autonomous mobile robots (AMRs) for parts transportation | WAREHOUSE & LOGISTICS |
| 12 | Bin-picking robot for handling a wide variety of unknown parts | WAREHOUSE & LOGISTICS |
| 13 | Advanced defect detection with robotic vision system | QUALITY |
| 14 | Autonomous maintenance inspection robots | MAINTAIN |
19 © Hi Ventures 2026 / hi.vc June 2026
Sources: Physical AI: Powering the New Age of Industrial Operations, WEF / BCG, 2025
## Agentic Organization
The agentic organization
AI infrastructure, model capabilities, and agentic systems advanced more in the last twelve months than many expected. A**s AI moves beyond generating content and begins executing work, the focus shifts from technology itself to its economic impact, **particularly in regions with significant productivity upside such as Latin America.
The result is the emergence of the agentic organization: a company designed around the complementary strengths of humans, AI agents, and intelligent machines. **The winners of the next decade may be those that learn to orchestrate these systems most effectively.**
20 © Hi Ventures 2026 / hi.vc June 2026
Inspired by source: The symbiotic organization, QuantumBlack, McKinsey, 2026
## Landscape
Hi on AI
02
## Reverse Stack
INVESTMENT THEMES
Hi Ventures identifies **four AI investment themes** to watch in Latin America
| Name | Description |
| --- | --- |
| AI-native finance | AI is rewriting how Latin Americans access and manage their money. From underwriting to wealth management, AI-native infrastructure is closing the inclusion gap faster than any other wave of fintech. |
| Intelligent commerce | As global commerce accelerates and supply chains reorganize around a new nearshoring paradigm, AI is enabling businesses to sell, ship, and scale across borders with unprecedented speed and precision. |
| Augmented human | From healthcare and diagnostics to wellbeing, education, and personal productivity; AI is amplifying human potential across every dimension. Startups are building the tools that make people healthier and capable. |
| Physical systems & infrastructure | AI is crossing the digital-physical barrier. Robotics, manufacturing automation, energy systems and smart infrastructure are the next frontier where software meets atoms and capital-intensive industries get their AI moment. |
Source: Hi Ventures analysis
## Four Themes Explorer
Four themes to follow in Latin America
| Index | Name |
| --- | --- |
| 01 | AI-native finance |
| 02 | Intelligent commerce |
| 03 | Augmented human |
| 04 | Physical systems & infrastructure |
Why this trend?
Startups to watch
| Label | Text |
| --- | --- |
| Digital rails already laid: | E-invoicing, mobile banking, and open banking APIs across the region mean the data infrastructure for AI-native finance exists. The intelligence layer is what's missing. |
| Credit for the unscored: | More than 70% of adults gained financial access in the last decade. AI can now underwrite segments that traditional scoring cannot reach. |
| SME operations: | Millions of businesses still manage finances manually. Autonomous agents for invoicing, cash flow, and compliance are a massive wedge. |
| Regulatory tailwind: | Open banking frameworks in Brazil, Mexico, and Colombia are turning every fintech into a potential AI-native platform. |
| Category | Co 1 Name | Co 1 Logo | Co 2 Name | Co 2 Logo | Co 3 Name | Co 3 Logo | Co 4 Name | Co 4 Logo |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Digital banking solutions | Mendel |  | tapi |  | Prometeo |  | 1erian |  |
| Payments / collection mgmt. | Shinkansen |  | Karta |  | ARQ |  | trace finance |  |
| Investments mgmt. | Fintual |  | Cicada |  | Nexa Finance |  | Simetrik |  |
| Fraud detection & credit underwriting | Menlo |  | niva |  | Liora |  | brick |  |
| Tax mgmt. automation | Cenit |  | camu |  | Cifrato |  | taxdown |  |
| Label | Text |
| --- | --- |
| WhatsApp-native commerce: | Latin America’s most natural commercial channel is already conversational. AI turns manual chat-selling into autonomous, scalable storefronts. |
| Marketplace data density: | Billions in sales volume across regional platforms generate the training data for AI tools that serve the long tail of merchants. |
| Logistics orchestration: | AI-powered routing and demand forecasting are solving the region's structural last-mile challenges at a fundamental level. |
| Cross-border transparency: | AI is the essential layer for managing the complexity of regional trade, from customs compliance to real-time supply chain tracking. |
| Category | Co 1 Name | Co 1 Logo | Co 2 Name | Co 2 Logo | Co 3 Name | Co 3 Logo | Co 4 Name | Co 4 Logo |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Supply chain optimization | Nuvocargo |  | BackChannel |  | Desteia |  | Birdie |  |
| Data analytics and cloud solutions | Senzai |  | Undercover |  | Diio |  | ZeroEval |  |
| Sales intelligence software | Vambe |  | PatagonAI |  | Galo |  | Darwing AI |  |
| E-commerce personalization | Trinio |  | Melian |  | VerveMarket |  | PrimeroAI |  |
| Environment sustainability | Satellites on Fire |  | Aravita |  | FieldData |  | Bemagro |  |
| Label | Text |
| --- | --- |
| Wellbeing at scale: | Mental health remains underserved. Digital channels lower the stigma barrier, and AI-powered tools reach populations that would never walk into a clinic. |
| New categories of care: | The opportunity in augmentation is vastly larger than automating what professionals already do. AI enables services that did not exist before. |
| Specialist scarcity: | One specialist per 5,000 people in most rural areas. AI diagnostic tools multiply the reach of existing health infrastructure. |
| Adaptive learning: | Education and workforce development face a scale problem. AI tutors that adapt in real time are showing 2-3x improvements in early deployments. |
| Category | Co 1 Name | Co 1 Logo | Co 2 Name | Co 2 Logo | Co 3 Name | Co 3 Logo | Co 4 Name | Co 4 Logo |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Legal tech and AI applications | Enter |  | Magnar |  | Lexter |  | PrimeroAI |  |
| HR Tech & Workforce enablement | Runa |  | Tako |  | Franq |  | Hunty |  |
| Healthcare operations & triage | Konko AI |  | Yana |  | Telepatia |  | Leona |  |
| Insurance industry applications | Hanlde |  | Azos |  | Darwin |  | Momento |  |
| Decision making & productivity & agents | Tess |  | Faces |  | Comp |  | Inner AI |  |
| Label | Text |
| --- | --- |
| Nearshoring pull: | $41B in Mexican foreign direct investments in 2025, 37% to manufacturing. Every new factory floor is a deployment surface for AI-native automation. |
| Robot density gap: | Mexico has 62 robots per 10,000 workers. South Korea has 1,220. That 20x gap will close, and the companies building the automation layer will capture it. |
| Greenfield advantage: | New facilities can be designed around AI from day one rather than retrofitting legacy systems. |
| Physical industries, data-rich: | Manufacturing, mining, and energy represent $800B+ of Latin America GDP with vast operational data and almost no AI running on it. |
| Category | Co 1 Name | Co 1 Logo | Co 2 Name | Co 2 Logo | Co 3 Name | Co 3 Logo | Co 4 Name | Co 4 Logo |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Industrial automation and Robotics | Strive Robotics |  | Robot.com |  | Aerial Loop |  | | |
| AI native energy & mining operations | Bruna |  | Splight |  | Solar |  | | |
| Predictive Maintenance and Asset Intelligence | Datamint |  | Fracttal |  | Aimirim |  | | |
© Hi Ventures 2026 / hi.vc June 2026
## Layered Stack
Beyond the horizon: 3 big things to expect in the next year
| Index | Heading |
| --- | --- |
| 1 | Latin America produces world-class AI enterprises |
| 2 | The first wave of agent-native startups |
| 3 | The next AI consumer platform emerges in the region |
| Label | Text |
| --- | --- |
| AI moves from experimentation to advantage: | The region's largest companies are shifting from pilots to deployment across core operations. |
| Local complexity becomes proprietary data: | Regulation, language, fragmentation, and informal markets create training data global competitors do not possess. |
| A new generation of AI leaders emerges: | The companies that master AI first will widen the gap with slower competitors. |
| Label | Text |
| --- | --- |
| No legacy, no migration: | Many startups are being built after the arrival of AI, avoiding decades of software and workflow baggage. |
| Complexity is fertile ground: | Fragmented industries, regulatory variation, and operational inefficiencies create ideal conditions for agent-driven businesses. |
| A generation of AI-native founders: | Thousands of entrepreneurs are building their first companies assuming agents are part of the team from day one. |
| Label | Text |
| --- | --- |
| Conversation becomes the operating system: | Consumers increasingly access services by talking to agents rather than navigating apps. |
| Messaging creates a unique advantage: | The region's messaging-first habits accelerate adoption of agent-based experiences. |
| Services come to the user: | AI reduces the friction of discovery, navigation, and execution. |
© Hi Ventures 2026 / hi.vc June 2026
## tess.im x h1 ad
Loved by over **16K** professionals
| Avatar |
| --- |
|  |
|  |
|  |
|  |
Scale with a Hybrid Workforce
Turn every employee into a manager of Human-AI agents. The future of work is here: hybrid squads across your entire org chart, with no user seat costs.
Keep vibing. Keep working.
Tess Cowork
Tess Corp
Hire
Structure
People
Searching the right personality
Search agents
Ask tess or assign a task...
Tess is thinking...
| ID | Name | Role | Caption | Badge | Status | Avatar |
| --- | --- | --- | --- | --- | --- | --- |
| john | John Doe | Workspace Member | Workspace owner and operator | | |  |
| tess | Tess | AI Orchestrator | Coordinates your agent team | | |  |
| director | Ethan Cole | Director of Research | Leads the research function | | working |  |
| trend | Olivia Park | Trend Spotter | Monitor emerging trends across markets | REVIEW | review |  |
| analyst | Maya Patel | Market Analyst | Analyze market dynamics and signals | | working |  |
| Name | Role | Caption | Prompt | Avatar |
| --- | --- | --- | --- | --- |
| Marcus Chen | Data Synthesizer | Compile insights into reports | Compile our research into clean reports |  |
| Emma Brooks | Outreach Agent | Draft and send follow-ups | Hire someone to draft and send the follow-up emails |  |
## Capital
Corporates
03
## Insights List
Corporate:
key insights
| Icon | Body |
| --- | --- |
| :icon-brain-circuit: | **71% of corporates have adopted AI **at the team or function level while **only 5% haven't explored it yet** |
| :icon-trending-up: | **Productivity gains remains the top measure of AI impact** but 25% of companies still don't have formal metrics in place |
| :icon-bot: | **Data privacy (29%) and lack of technical talent (28%) are the top barriers**, followed by AI evolving too fast (24%), a new concern not present in 2025 |
| :icon-users: | **58% of corporates are at least piloting agents **and only 12% have no plans to deploy (vs. 43% piloting and 18% no plans in 2025) |
## Expert Insight
MERCADO LIBRE
**Rewiring at scale:** insights from Leandro Cuccioli
“
The biggest barrier is cultural, not technical. For the first time, anyone can build their own tools.
Leandro Cuccioli
SENIOR VP, MERCADO LIBRE
Source: Hi Ventures interview with Leandro Cuccioli on Hi Perspectives
| Heading | Body |
| --- | --- |
| One of the world's largest AI rollouts | **Mercado Libre deployed Claude for Work to 31,000 employees, one of the largest rollouts in the world to date.** With more than 95 percent of the workforce using AI, the company is now rethinking what talent looks like. The requirement is no longer a technical background but the ability to reason and connect dots. |
| Where the impact shows first | With the business three times larger than three years ago, AI now handles most customer interactions, with humans focusing on complex escalations. On the developer side, more than 18,000 engineers use AI coding tools across 47,000 applications. **Financial planning, tax, and procurement were the first business areas to transform** because they had the most fragmented data and manual work, exactly where agents deliver immediate value. |
| The real moat is in the data | Mercado Libre treats foundation models as interchangeable, and is building a middleware layer that routes tasks to the most efficient model, where tokens can cost a tenth of the price. The conviction guiding this work is that **the real advantage is not which model to use but how to orchestrate them with proprietary data**. That orchestration is the path the company is on, and one it keeps building. |
© Hi Ventures 2026 / hi.vc June 2026
## Market Synthesis
Most companies have adopted AI at the team or function level
Which of the following best describes your organization's current stage in adopting AI?
| Label | Value |
| --- | --- |
| Team/function-level adoption | 71 |
| Individual productivity only | 23 |
| Not used at all | 5 |
| Percent | Description |
| --- | --- |
| 71% | of the companies have AI at the **team or function level** — specific teams have adopted AI tools in their work. |
| 23% | of the companies are at **individual productivity only** — employees use tools like ChatGPT ad-hoc but no team-level adoption. |
|
|
|
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
## Bar Chart
Leaders say efficiency gains are what interests them most about AI’s impact
Improving productivity and efficiency is the top interest for corporate leaders (46%), well ahead of making better decisions (21%) and offering more personalized services (17%).
What interests corporate leaders most about AI?
Source: Hi Ventures State of AI Survey 2026, n=167 | Corporate respondents
| Label | Value (%) |
| --- | --- |
| Improving productivity and efficiency | 46 |
| Making better decisions | 21 |
| Offering more personalized services | 17 |
| Reaching new customers or markets | 13 |
| Other | 2 |
| None | 2 |
| Stat | Text |
| --- | --- |
| 46% | say **productivity & efficiency.** |
| 2% | Only 2% say **none**. Virtually every corporate leader sees something to be excited about. |
## Line Graph
Corporate AI agent adoption is broad but cautious
as most are still piloting
Is your organization currently using or planning to deploy AI agents?
When it comes to use cases for agents, **customer service automation is the top among corporates (32%),** followed by report generation and analytics (16%).
Yet in contrast to startups, where 53% already run agents in production, **corporates are moving more cautiously.**
Sources: Hi Ventures - State of AI survey 2026, n=98 corporates
| Label | Value (%) |
| --- | --- |
| No current plans | 12 |
| Yes, agents are deployed in production | 30 |
| Yes, piloting agents in limited deployments | 29 |
| Planning to deploy within the next 12 months | 13 |
| Exploring the possibility | 16 |
## ROI Metrics
Efficiency sets the ROI bar in Latin America
How does your company currently measure the impact or ROI of AI initiatives?
| Label | Value |
| --- | --- |
| Productivity gains (time saved, faster delivery) | 35 |
| Cost reduction | 15 |
| Operational metrics specific to AI use cases | 6 |
| Revenue growth | 4 |
| Quality of AI outputs | 4 |
| Enhanced decision-making | 4 |
| No formal metrics in place yet | 24 |
Productivity tops AI success metrics in Latin America (35%) while globally, innovation leads at 64% and cost improvement at 38% (McKinsey/Stanford HAI 2026). In Latin America, cost reduction ranks second (15%), underscoring **the regional focus on operational efficiency**, while globally, organic revenue growth is cited by 33% of organizations.
Yet, 24% still don't have formal metrics in place.
Source: Hi Ventures - State of AI survey 2026, Stanford HAI - AI Index report 2026
35 © Hi Ventures 2026 / hi.vc June 2026
## Key Metrics
However, resource constraints and technical capability gaps remain the top barriers to corporate AI adoption
Why hasn't your organization explored AI yet?
| Value | Label |
| --- | --- |
| 22 | Limited time or resources to prioritize AI |
| 20 | Lack of technical expertise or internal capabilities |
| 15 | Workforce concerns or organizational resistance |
| 14 | Concerns about data privacy or security |
| 12 | Unclear ROI or business case for AI |
| 3 | Waiting for proven use cases in our industry |
As surfaced in the interviews, **bridging the technical gap and elevating AI as a cross-functional priority beyond tech teams** are the levers that move companies from no adoption to pilots, and from pilots to scaling.
Globally, **knowledge and training gaps remain the top obstacle to scaling AI**, cited by 59% of organizations — up from 51% in 2024 (McKinsey/Stanford HAI 2026).
Source: Hi Ventures - State of AI survey 2026, n=65 corporates (excl. those who already adopted AI broadly). McKinsey & Company Survey 2025 via Stanford HAI AI Index 2026
30 © Hi Ventures 2026 / hi.vc June 2026
## Data Breakdown
Corporate AI adoption is broad but shallow
For each function, how deeply is AI adopted?
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
| Label |
| --- |
| Embedded |
| Broad |
| Regular |
| Early |
| None |
| N/A |
| Function | Embedded | Broad | Regular | Early | None | N/A |
| --- | --- | --- | --- | --- | --- | --- |
| Data, analytics & BI | 21 | 27 | 19 | 20 | 0 | 9 |
| Engineering & software dev | 19 | 22 | 18 | 16 | 14 | 9 |
| Product mgmt & design | 14 | 11 | 20 | 26 | 18 | 10 |
| Strategy & leadership | 12 | 16 | 27 | 26 | 0 | 15 |
| Operations & admin | 11 | 14 | 26 | 27 | 0 | 18 |
| Customer support | 10 | 18 | 21 | 24 | 20 | 5 |
| Sales, Marketing & growth | 8 | 19 | 23 | 31 | 0 | 13 |
| Stat | Text |
| --- | --- |
| 21% | Of corporates have AI fully embedded in Data & BI — the deepest adoption of any function |
| 14% | Average None across functions — showing significant room for deeper adoption vs startups where Engineering alone hits 62% embedded |
## Data Snapshot
Only one in four formally evaluate employees on AI usage
Do you evaluate how employees use AI in their work?
| Label | Value | Color |
| --- | --- | --- |
| Informally | 33 | var(--color-2) |
| No | 24 | var(--color-5) |
| Yes — most or all roles | 17 | var(--color-4) |
| Planning to (next 12 months) | 17 | var(--color-1) |
| Yes — AI roles only | 8 | var(--color-3) |
| Stat | Text |
| --- | --- |
| 33% | assess AI usage informally — **managers consider it but it is not part of formal performance reviews.** |
| 17% | plan to start evaluating AI usage within the next 12 months — **suggesting formal assessment is gaining momentum.** |
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
## Key Metrics
User trust, evaluation, and cost are the top challenges corporates face when deploying AI agents
What is the biggest technical challenge when building/deploying AI agents?
| Value | Label |
| --- | --- |
| 22 | User trust & adoption |
| 20 | Evaluation & quality measurement |
| 19 | Cost per agent run at scale |
| 18 | Context / memory management |
| 16 | Multi-agent coordination |
| 15 | Reliability |
| 12 | Tool use & API integration |
| 10 | Observability & debugging |
The challenges are **remarkably evenly distributed,** no single issue dominates. This suggests that deploying AI agents in corporate environments is **hard across the board**, from trust to cost to reliability.
Notably, **user trust & adoption leads over purely technical issues**, signaling that the **human side of agent deployment** is as critical as the engineering.
Source: Hi Ventures - State of AI survey 2026, n=98 corporates (multi-choice)
33 © Hi Ventures 2026 / hi.vc June 2026
## Deep Dive
Most corporates have or are developing a formal AI governance framework
Do you have a formal AI governance framework?
| Label | Value | Highlight |
| --- | --- | --- |
| Yes — formal framework | 33 | No |
| In progress | 45 | No |
| No formal framework | 22 | Yes |
| Stat | Text |
| --- | --- |
| 78% | of corporates have or are developing a **formal AI governance framework.** |
| 45% | say their governance framework is **still in progress**, signaling growing but incomplete oversight. |
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
© Hi Ventures 2026 / hi.vc June 2026
34
## Data Snapshot
Human-in-the-loop is the dominant autonomy model
What level of autonomy do your AI agents currently operate at?
| Label | Value | Color |
| --- | --- | --- |
| Fully autonomous | 5 | var(--color-3) |
| Human-in-the-loop | 40 | var(--color-2) |
| Not yet | 29 | var(--color-5) |
| Suggestive | 14 | var(--color-4) |
| Bounded autonomy | 12 | var(--color-1) |
| Stat | Text |
| --- | --- |
| 40% | are piloting agents with human-in-the-loop oversight — **humans review and approve each action.** |
| 5% | Only have reached fully autonomous agent deployment, **while 29% don't use AI agents yet.** |
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
## Deep Dive
Also, corporates consider AI fluency when recruiting, but most do it informally
Are you assessing AI fluency when recruiting?
| Label | Value | Highlight |
| --- | --- | --- |
| Yes — all roles | 15 | No |
| Yes — tech roles only | 19 | No |
| Informally | 26 | No |
| Planning to | 18 | Yes |
| No | 21 | Yes |
| Stat | Text |
| --- | --- |
| 60% | of corporates consider **AI fluency when recruiting,** formally or informally. |
| 18% | are **planning to start assessing AI fluency** in the next 12 months. |
Source: Hi Ventures - State of AI survey 2026, n=98 corporates
© Hi Ventures 2026 / hi.vc June 2026
36
## Expert Insight
THE FUTURE OF COMMERCE
**The future of commerce:** Insights from Gonzalo Oyanedel
“
We're installing 2,300 robots. A window into the future of the human-machine relationship.
Gonzalo Oyanedel
HEAD OF AI, WALMART
Source: Hi Ventures interview with Gonzalo Oyanedel
| Heading | Body |
| --- | --- |
| Agentic commerce | Walmart is moving early in agentic commerce. After standalone industry approaches proved insufficient, Walmart partnered with OpenAI and Gemini to define AI-powered shopping. **The conviction: the best agentic experiences keep the entire transaction within a trusted environment.** Sparky, the shopping agent, will be scaling in LatAm through 2027. |
| How Walmart runs AI: people led, tech powered | Walmart organizes AI across capability building, agentic acceleration, and tailor-made AI/ML solutions. Last year it delivered over 8,000 training interactions, and today **roughly 75% of corporate employees actively use AI.** The key has been a cultural transformation program called "IAquí, IAhora," with four super-agents serving customers, sellers, associates, and developers. |
| Chile as a global hub | Walmart treats Chile as a global testbed: small enough to be low-risk, yet complex enough to test everything. It leverages diverse users, strong logistics, the world's third-fastest internet, and top talent at Latin American cost. **Consumer bets like Carrito Listo and Sparky were piloted in Chile first**, then exported worldwide. |
© Hi Ventures 2026 / hi.vc June 2026
## Expert Insight
AGENTS WITH VOICE
**Agents with voice:** insights from Sasha Glatt
“
Today companies are designing how their voice sound.
Sasha Glatt
GROWTH, ELEVENLABS
Source: Hi Ventures interview with Sasha Glatt
| Heading | Body |
| --- | --- |
| Voice is becoming a brand asset | Voice is becoming a brand asset at the same level as a logo or a color palette. Companies are deciding how their brand sounds across every customer interaction. ElevenLabs already works with 70 languages and plugs into Zendesk, Twilio, and Salesforce, so companies can add a voice layer on top of what they already have. **The shift is from text-based bots that feel generic to voice agents that carry the personality of the brand.** |
| Beyond customer support | Finance, healthcare, and government services turned out to be the strongest proving grounds for voice AI. The common thread is that the **highest-impact conversations are not the obvious customer support ones but the ones where tone and empathy change the outcome.** In healthcare, tone calibration by medical specialty is already a differentiator. Also in financial services in tasks like collections because the agent never loses patience and can adapt to the emotional weight of the clients. |
| Why LatAm is a natural fit | Mexico and Brazil are among the world's highest markets for WhatsApp voice notes. People in the region already prefer speaking over typing, which makes voice AI adoption feel less like a technology shift and more like a natural extension. **Agents with local accents and local vocabulary are what makes adoption work.** |
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Incode** is an AI-powered identity platform for global enterprises. Its **Agentic Identity** capability verifies which human is behind an AI agent, what they authorized, and what the agent is allowed to do.
Agentic identity:
after more than 7 billion identity checks, Incode is now piloting identity for agents to bring trust and accountability to the agent economy
| Index | Body |
| --- | --- |
| 01 | **AI-enabled fraud grew 1,210% in 2025 as agents can now log into accounts**, fill forms, contact support, initiate transactions, and test identity controls at machine scale.
**The new identity gap is knowing who controls the agent**, what they authorized, and whether the action is within scope. |
| 02 | **Incode built Agentic Identity**, a layer that connects agents to verified human identity: it **detects agent activity** using behavioral and device signals, **binds agents to verified humans** when access is granted, and **scopes permissions and triggers step-up verification** when risk changes.
Every agent action remains traceable to a human identity and permission layer. |
| 03 | Every agent action traces back to a verified human, creating a full audit trail of who authorized what. High-risk actions require live confirmation, reducing fraud exposure without blocking legitimate users.
Pilots with leading financial institutions in the US and Latin America show that verified agent access reduce friction.
|
39
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Runa** built Latin America's leading payroll platform, serving more than 3,500 companies across five countries. Now they're using AI agents to tackle one of the region's toughest workforce development problems.
AI training agents:
Runa built an AI agent that can train a new hire on any scenario before they ever face a real customer
| Index | Body |
| --- | --- |
| 01 | Latin America has one of the **highest employee turnover rates** in the world. Companies with thousands of frontline workers need to onboard and train new staff constantly, but traditional coaching depends on managers with limited time.
Feedback varies between trainers, and scaling in-person sessions across multiple locations **is slow and expensive.** |
| 02 | Runa built an AI Roleplay agent turning employee's phone into a training coach:
• The **AI creates realistic practice simulations** from a simple text.
• Employees **practice through voice** in any language getting scored.
• Companies build **structured learning paths by role** from beginner to certification.
The platform turns workforce training from a scheduling problem into a scalable system. |
| 03 | Dentalia, a dental clinic chain that **deployed AI training across all locations** to standardize patient care.
Training **runs 24/7 in any language**, removing geographic and scheduling barriers.
More than **3,500 payroll clients across five countries now use Runa's platform** to develop their teams. |
40
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Mendel** is Latin America's leading AI-powered spend management platform. It uses AI agents to automate expense tracking, invoice recovery, corporate travel, and tax compliance for large companies across Mexico, Argentina, and Chile.
AI-powered spend management:
Mendel's AI agents save finance teams 150 hours of admin work a month and cut non-deductible expenses by 20%
| Index | Body |
| --- | --- |
| 01 | Large enterprises manage thousands of transactions across employees, suppliers, and operations. Most expense tracking is still manual, relying on spreadsheets, paper receipts, and cash advances with no real-time visibility.
**When expenses are paid without proper invoicing, non-deductibility rates reach up to 70% and companies end up losing money.** |
| 02 | Mendel built an AI-powered spend platform that replaces manual processes end to end:
• **Smart cards** with real-time budget controls and restrictions.
• **AI invoice recovery** with photo of the ticket that retrieves the full CFDI.
• **Automated invoice** validation checking every expense against the SAT and syncs with the ERP.
• **AI agents** handle reconciliation, anomaly detection, and compliance process. |
| 03 | **150 hours saved** per month on administrative tasks across client teams.
Over **US$20,000 in admin costs eliminated** monthly per client.
**SOC 2 compliant, with native ERP and bank integrations**
Trusted by **Mercado Libre, FEMSA, McDonald's, OXXO, KPMG, Suzuki,** and Viva Aerobus. |
41
© Hi Ventures 2026 / hi.vc June 2026
## Interlude
"You will see that over the next three years, everything that can be simplified and made easier, particularly repetitive processes or those with a high probability of human error, should not remain manual. They will all be driven by AI"
Juan Manuel Valle
CEO Afore Coppel
## Cosmo Add
PRODUCTION-READY CAMPAIGNS IN MINTUES
Discover a universe inside your brand
JOIN THE WAITLIST
https://cosmo.senzai.ai/
## Infrastructure
Startups
04
## Startups Insights
Startups:
key insights
| Icon | Body |
| --- | --- |
| :icon-rocket: | **99%** of startups use **AI at least internally**. |
| :icon-brain-circuit: | **78% of startups have AI embedded in their core product** (59% AI-native, 19% have shipped AI features to customers) |
| :icon-package-open: | **Claude is used by 83% of startups**, overtaking OpenAI (66%), while Claude Code leads AI coding tools at 69% |
| :icon-bot: | **72% of startups are at least piloting agents. **Only 9% have no plans (vs. 62% and 10% in 2025) |
## Expert Insight
THE GREAT EQUALIZER
The great equalizer: insights from Paqui Casanueva
“
AI is becoming the operating system of the economy. **There is no other option left but to embrace it.**
Paqui Casanueva
BOARD PRESIDENT, ENDEAVOR
Source: Hi Ventures interview with Paqui Casanueva on Hi Perspectives
| Heading | Body |
| --- | --- |
| AI is a new layer on top of everything | Everyone assumes AI will dehumanize work, but the opposite is happening. It removes repetitive, low-value tasks and moves people from transactional roles to strategic ones. **The data was always there as raw material, but now we finally have the machinery to extract value from it.** That shift is what changes how teams operate, how decisions get made, and where humans actually add value. |
| AI lowers the cost of ambition | Every generation of entrepreneurs benefits from a new technological shift. For this generation, it is AI. **Founders can now build products faster, automate work that once required entire teams, and reach customers with far fewer resources than ever before.** The best entrepreneurs are not building around AI itself, they are using AI to solve meaningful problems. |
| Mexico's complexity is its advantage | A microentrepreneur with a phone now has computing power comparable to what a large corporation had five years ago. AI is leveling the field. Mexico has 40 million people without a single financial service and 40 million relatives in the United States, a double market waiting to be unlocked. **If you can solve a problem in Mexico, with all its complexity, your model works anywhere in Latin America.** |
© Hi Ventures 2026 / hi.vc June 2026
## Startups AI Embed
STARTUP ADOPTION
Majority of startups embed AI in their core product and use AI at least internally
Most startups are building AI into what they sell, not just how they operate.
Is AI embedded in your core product?
Source: Hi Ventures - State of AI survey 2026, n=268 startups
| Label | Value (%) |
| --- | --- |
| Product is AI-native | 59 |
| AI capabilities shipped | 19 |
| AI features in dev / planned | 14 |
| No AI in product | 8 |
| Stat | Text |
| --- | --- |
| 78% | of startups have **AI embedded** in their core product. |
| 31% | of the startups are in **B2B SaaS / Enterprise Software**, the largest sector. |
| 62% | have teams with **10 people or less.** |
|
|
|
## Startups Bottlenecks
Talent, reliability, and compute cost are the top bottlenecks startups face with AI
What are the biggest bottlenecks to deploying AI?
Multiple choice question
| Label | Value (%) | Highlight |
| --- | --- | --- |
| Lack of technical talent | 23 | Yes |
| AI agent reliability in production | 23 | No |
| Compute cost | 22 | No |
| Lack of bandwidth | 19 | No |
| High implementation costs | 18 | No |
| Data privacy | 16 | No |
| Lack of internal investment | 13 | No |
| Difficulty proving value | 10 | No |
| Regulatory uncertainty | 8 | No |
| Lack of commitment | 6 | No |
The top 3 bottlenecks are in a **virtual tie: talent (23%), agent reliability (23%), and compute cost (22%)** — reflecting that AI deployment challenges are structural, not singular.
Notably, **6% cite lack of commitment** as a bottleneck, suggesting a segment of startups has already overcome.
Source: Hi Ventures - State of AI survey 2026, n=268 startups (multi-choice)
48
© Hi Ventures 2026 / hi.vc June 2026
## Model Preferences
AI model preferences among startups (2023–2026)
Which AI models does your startup use?
| Name | Color | 2023 | 2024 | 2025 | 2026 |
| --- | --- | --- | --- | --- | --- |
| ChatGPT | #58BED6 | 92 | 97 | 89 | 66 |
| Claude | #000000 | | 29 | 45 | 83 |
| Gemini | #ABC9CE | | 26 | 48 | 59 |
| DeepSeek | #50C8BA | | | 19 | 10 |
| Label |
| --- |
| 2023 |
| 2024 |
| 2025 |
| 2026 |
83%
Of startups now use Claude, up from 29% in 2024 — a 2.9x increase in just two years, overtaking OpenAI as the #1 model among Latin American startups
66%
Of startups still use OpenAI models, down from 97% in 2024, as startups shift from single-model reliance to multi-model experimentation across Claude, Gemini, and emerging players
Source: Hi Ventures — State of AI Survey 2023, 2024, 2025, 2026.
55 © Hi Ventures 2026 / hi.vc June 2026
## Startups Budget
Half of startups allocate more than 10% of their operating budget to AI
What percentage of your operating budget is allocated to AI?
| Label | Value (%) |
| --- | --- |
| Less than 1% | 9 |
| 1–5% | 16 |
| 5–10% | 16 |
| 10–25% | 20 |
| 25–50% | 15 |
| More than 50% | 16 |
| Don't know | 8 |
| Stat | Text |
| --- | --- |
| 16% | of startups allocate more than 50% of their budget to AI, reflecting the rise of AI-native business models.
|
| 8% | don't track AI spending separately. As AI becomes embedded across functions, isolating its cost may become more difficult.
|
Source: Hi Ventures - State of AI survey 2026, n=268 startups
50
© Hi Ventures 2026 / hi.vc June 2026
## Startups Risks
Both Latin American startups and global corporates see hallucination & inaccuracy as the main risk to deploying AI
What are the biggest risks to deploying AI?
Multiple choice question
| Label | Value (%) |
| --- | --- |
| Inaccuracy and hallucination | 58 |
| Data privacy breaches | 39 |
| Cyber-security vulnerabilities | 33 |
| Regulatory compliance failure | 20 |
| Explainability and auditability issues | 19 |
| Workforce displacement | 10 |
| IP and copyright infringement | 8 |
| I don't see risks | 7 |
| Other | 9 |
Latin America startups vs. Global corporates
| Stat | Text |
| --- | --- |
| | Inaccuracy & hallucination leads in Latin America **(58%)** and globally **(74%)**. |
| | Only **33%** of Latin American startups list cybersecurity as a top AI risk vs. **72%** of global organizations. |
Source: Hi Ventures - State of AI survey 2026, n=268 startups. McKinsey & Company Survey 2025 via Stanford HAI AI Index 2026
51
© Hi Ventures 2026 / hi.vc June 2026
## Startups Platforms
Anthropic Agents SDK leads agent platforms, followed by n8n and LangChain
Which AI agent/automation platforms does your company use?
Multiple choice question
| Label | Value (%) |
| --- | --- |
| Anthropic Agents SDK | 29 |
| n8n | 23 |
| LangChain / LangGraph | 22 |
| OpenAI Assistants API | 13 |
| ElevenLabs | 10 |
| Amazon Bedrock Agents | 9 |
| Vertex AI Agent Builder | 7 |
| Microsoft Copilot Studio | 5 |
| Zapier AI | 4 |
| Make | 3 |
The agent platform landscape is **highly fragmented** — the top tool (Anthropic Agents SDK) captures only 29%, and **19% say 'None'**, signaling many startups build custom solutions.
Open-source and low-code platforms (**n8n, LangChain, Make, Zapier**) collectively represent strong adoption, reflecting a **build-it-yourself culture** among LatAm startups.
Source: Hi Ventures - State of AI survey 2026, n=268 startups (multi-choice)
52
© Hi Ventures 2026 / hi.vc June 2026
## Startups Embedded
AI adoption runs deep in technical functions but, but back-office functions remain in early stages
For each function, how deeply is AI adopted?
Source: Hi Ventures — State of AI survey 2026, n=265 startups
| Label |
| --- |
| Embedded |
| Broad |
| Regular |
| Early |
| None |
| Function | Embedded | Broad | Regular | Early | None |
| --- | --- | --- | --- | --- | --- |
| Engineering & software dev | 62 | 17 | 8 | 8 | 5 |
| Product mgmt & design | 41 | 28 | 15 | 11 | 5 |
| Data, analytics & BI | 44 | 23 | 14 | 12 | 7 |
| Strategy & leadership | 26 | 25 | 22 | 13 | 12 |
| Sales, Marketing & growth | 21 | 29 | 25 | 17 | 8 |
| Customer support | 22 | 24 | 23 | 18 | 11 |
| Operations & admin | 15 | 26 | 26 | 20 | 10 |
| Stat | Text |
| --- | --- |
| 62% | Of startups have AI fully embedded in **Engineering** — the deepest adoption of any function |
| 15% | Have AI embedded in **Operations & Admin** — the shallowest adoption, with 10% not using AI at all |
## Startups Agents
Most startups are at least piloting AI agents and only few don't have plans to deploy it
Is your organization currently using or planning to deploy AI agents?
| Label | Value | Color |
| --- | --- | --- |
| No current plans | 9 | #abc9ce |
| Yes, already using agents in production | 53 | var(--color-1) |
| Yes, piloting agents | 19 | var(--color-5) |
| Yes, planning to test within the next 12 months | 13 | var(--color-4) |
| Exploring the possibility | 7 | var(--color-2) |
More than half of startups have moved past experimentation and into production-grade agent deployment.
The competitive **learning curve is now compounding** as each deployment cycle generates proprietary data, refined prompts, and operational know-how that newcomers can't shortcut.
Source: Hi Ventures - State of AI survey 2026, n=268 startups
55
## Startups Use Cases
Research and workflow coordination lead startup agent use cases
What are your primary AI agent use cases?
Multiple choice question
| Label | Value (%) | Highlight |
| --- | --- | --- |
| Research, analysis & knowledge work | 48 | yes |
| Workflow coordination across apps or systems | 47 | no |
| Coding or dev-ops | 44 | no |
| Customer service automation | 41 | no |
| Sales automation | 29 | no |
| Voice or conversational agents | 29 | no |
| Marketing execution | 23 | no |
| Other | 10 | no |
Top four use cases are tightly clustered between **41% and 48%** — no single dominant application. **Research & workflow coordination (48% and 47%)** lead, followed closely by **coding/dev-ops (44%)** and **customer service (41%)**.
Unlike corporates, where customer service leads at **32%**, startups spread agent use more evenly across functions.
Source: Hi Ventures - State of AI survey 2026, n=268 startups (multi-choice)
56
© Hi Ventures 2026 / hi.vc June 2026
## Startups Autonomy
Human-in-the-loop dominates. Startups keep humans reviewing every agent action
What level of autonomy do your AI agents currently operate at?
| Label | Value (%) |
| --- | --- |
| Suggestive | 13 |
| Human-in-the-loop | 47 |
| Bounded autonomy | 24 |
| Fully autonomous | 16 |
| Stat | Text |
| --- | --- |
| 47% | of startups using agents keep them in **human-in-the-loop mode** — humans review and approve each action |
| 16% | have reached **fully autonomous agent deployment**, compared to only 5% of corporates |
Source: Hi Ventures - State of AI survey 2026, n=225 startups (those using or piloting agents)
57
© Hi Ventures 2026 / hi.vc June 2026
## Startups Coding Tools
Claude Code leads AI coding tools at 69%, with Anthropic tools dominating the developer stack
Which AI coding/developer productivity tools does your company use?
Multiple choice question
| Label | Value (%) |
| --- | --- |
| Claude Code | 69 |
| Claude Cowork | 43 |
| Cursor | 34 |
| Codex | 31 |
| GitHub Copilot | 22 |
| Lovable | 16 |
| Google Gemini Code Assist | 14 |
| v0 by Vercel | 6 |
Startups show strong Anthropic tool adoption: **Claude Code and Claude Cowork** are the top two coding tools, ahead of **Cursor and Codex**.
**GitHub Copilot** sits at 22%, behind four AI-native coding tools.
Source: Hi Ventures - State of AI survey 2026, n=268 startups (multi-choice)
58
© Hi Ventures 2026 / hi.vc June 2026
## Startups Challenges
Context management, evaluation, and cost are the top challenges startups face deploying AI agents
What is the biggest technical challenge when building/deploying AI agents?
| Label | Value (%) |
| --- | --- |
| Context / memory management | 35 |
| Evaluation & quality measurement | 32 |
| Cost per agent run at scale | 32 |
| Reliability | 31 |
| User trust & adoption | 28 |
| Observability & debugging | 21 |
| Multi-agent coordination | 19 |
| Tool use & API integration | 12 |
| Stat | Text |
| --- | --- |
| | Unlike corporates where user trust leads (22%), startups' top challenge is **context/memory management, **a more technical, implementation-level concern |
| | The top 4 challenges are within 4 points of each other (35%–31%), confirming that **agent deployment is hard across the board** |
Source: Hi Ventures - State of AI survey 2026, n=268 startups (multi-choice)
59
© Hi Ventures 2026 / hi.vc June 2026
## Startups Capabilities
Most startups have shipped process & analyze data capabilities
How deeply are these AI capabilities integrated into your product?
Source: Hi Ventures — State of AI survey 2026, n=268 startups
| Label |
| --- |
| Core |
| Foundational |
| In production |
| Piloting |
| Not offered |
| Capability | Core | Foundational | In production | Piloting | Not offered |
| --- | --- | --- | --- | --- | --- |
| Process & analyze data | 13 | 46 | 20 | 11 | 10 |
| Conversational interfaces | 18 | 35 | 15 | 15 | 11 |
| Personalize & recommend | 16 | 34 | 15 | 15 | 20 |
| Autonomous actions | 15 | 33 | 12 | 15 | 25 |
| Generate content | 14 | 31 | 9 | 15 | 31 |
| Stat | Text |
| --- | --- |
| 80% | Of startups have shipped "Process & analyze data" capabilities — the most mature AI use case |
| 55% | Have shipped "Generate content" — the least mature, with 21% not even offering it |
## Expert Insight
THE FUTURE OF AGENTS
**The future of agents:** insights from Ricardo Barros
“
If an AI company is not global from day one, **it does not exist**
Ricardo Barros
CO-FOUNDER & CEO, TESS AI
Source: Hi Ventures interview with Ricardo Barros
| Heading | Body |
| --- | --- |
| From chatbot to AI employees | Most enterprise AI is chat-based, confined to tech teams, and hidden behind command-line interfaces. **The three main barriers to fix for this are security fears over what agents can touch, a UX built for engineers, and the assumption that automation means fewer jobs.** Until those are solved, agentic platforms stay below the average adoption at the company level. |
| The bottom-up model that breaks the pattern | Tess does not deploy top-down. **Employees create and share AI employees themselves through a visual interface,** no code or IT approval needed. A permission layer gives enterprises control, and multi-model orchestration routes each task to the best option across 300 models. The result are 16,000 users and 600,000 autonomous tasks growing inside organizations. |
| The adopters nobody expected | Sales and marketing teams saw returns first, followed by operations. But the real surprise is what is coming next on manufacturing plants, infrastructure companies, pharma, and publicly listed firms dealing with regulatory compliance. **High-volume, repetitive workflows where the agent does not replace a person but replaces the tools that person was toggling between will be the next step.** |
© Hi Ventures 2026 / hi.vc June 2026
## Expert Insight
FULL STACK VIEW
**Full stack view:** insights from Sandra Daza
“
Every company knows what hurts most. **AI just made it possible to finally fix it.**
Sandra Daza
SENIOR DIRECTOR OF AI AND CLOUD, ORACLE
Source: Hi Ventures interview with Sandra Daza on Hi Perspectives
| Heading | Body |
| --- | --- |
| AI democratizes advanced computing | Oracle moves across every level of the technology stack, from infrastructure and databases to integration and business applications. Today, a new layer has now opened up with AI services including generative models, computer vision, translation, and speech. **These sit on top of existing infrastructure and are available to anyone, from the largest bank in Latin America to a one-person startup.** |
| AI rewards organizational change | As much as 90 percent of enterprises are investing in AI today. However, most companies ran a proof of concept last year and stopped there. AI does not work if you plug an agent into a broken process. It works when you redesign the process defining where a human must stay in the loop and bring people along through cultural change. **The CEO who asks what has always hurt the most, and builds from there, is the one who sees results.** |
| Latin America is adopting faster than people think | **Latin America is outpacing the United States in consumption of certain LLM-enabling services.** This is the first industrial revolution that does not require massive capital investment. The region has talent, fast adoption culture, and world-class institutions. The barrier is not technology anymore. |
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Wonder Brands** is one of Latin America's leading e-commerce platforms and the #1 professional merchant on Mercado Libre®. **Coffee** is its AI-agent platform layer that predicts demand and recommends purchases for daily execution.
AI-agents layer:
Coffee cut forecast error from 30 percent to single digits and halved inventory coverage to under 45 days
| Index | Body |
| --- | --- |
| 01 | WB operates a **high-volume** commerce business, managing thousands of SKUs across brands, channels and categories from dozens of suppliers.
As businesses grow, planning purchases and inventory becomes increasingly **manual and fragmented** so buyers have to reconcile demand, MOQs, lead times and cash flows before deciding what to buy. |
| 02 | Coffee shifted the buyer role from calculating decisions to validating recommendations.
WB built **Coffee**, an AI-agent platform layer that compressed planning cycle from weekly to daily:
• **More than 1,300 SKUs** across 7 brands planned daily
• Generating **365-day projections**
• Incorporating MOQ, lead time and stock coverage agentic process |
| 03 | **US$3M in purchases** /mo generated by agents and validated by humans through the **Purchase Tool**
**Halved days of coverage** freeing up about half the working capital tied in inventory
Weekly cash flow **reporting became daily** AI-generated with visibility into KPIs and operating priorities |
63
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Klarna** serves 35 million customers and is one of the world's leading buy-now-pay-later platforms. They deployed **ElevenLabs Agents** as first-line voice support for instant customer service.
AI-powered support:
Klarna cut time to resolution by ten times serving 35 million US customers with ElevenLabs Agents as first-line phone support
| Index | Body |
| --- | --- |
| 01 | Klarna handles a high volume of phone calls, most of them informational, **asking for payment status, product guidance or understanding next steps.**
These simple requests still required queueing for a human agent, **increasing wait times and operational load** while pulling support teams from complex cases. |
| 02 | Klarna deployed **ElevenLabs Agents** as the first touchpoint for US phone support:
• **Low-latency** voice features at scale.
• **Automatic escalation to human** agents when complexity is detected.
• **Enterprise-grade security** and data controls aligned with regulations.
The AI can now handle the friction without compromising the relationship with clients. |
| 03 | Resolution times **reduced by ten times** for queries handled by the agent.
**Live for 35 million US customers** with instant response.
**Improved customer experience** and operational efficiency.
Now **scaling to serve all customers** worldwide using this service. |
64
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Cashea** is Venezuela's leading buy-now-pay-later platform, expanding access to consumer credit through a modern financial infrastructure. **Cheo** is its AI-agent for customer operations, starting with support and expanding toward collections.
AI-agent layer:
Cheo resolves up to 60% of Cashea's customer conversations while helping the company scale credit without increasing headcount
| Index | Body |
| --- | --- |
| 01 | Cashea operates in a market where traditional consumer credit infrastructure had largely disappeared. **There is no credit bureau, so the company had to build its own risk intelligence.**
As the portfolio scaled, **new users brought more payment validations, collections touchpoints, and product journeys** that would require more human operators. |
| 02 | Cashea built **Cheo**, an AI conversational agent acting as the first layer for customers:
• **Resolves payment and product questions** across customer support.
• **Validates payments** and automates repetitive customer flows.
• **Prioritizes collections** by identifying users requiring intervention.
Cashea's proprietary **risk models learn from first-party repayment data** generated directly on its platform. |
| 03 | Cheo already **resolves half of customer conversations with 85% satisfaction.**
Better support protects repayment behavior by **reducing payment friction.**
**Automated conversations reveal recurring user activity.**
The long-term vision is a **financial agent that can anticipate user needs.** |
65
© Hi Ventures 2026 / hi.vc June 2026
## Interlude
"The future of AI is not intelligence. It’s trust. Intelligence gets attention. Trust earns a place in people’s lives”
Andrea Campos
Founder Yana
## Explore Faces
https://faces.app/
## Outlook
Venture Capital
05
## Year in Review Timeline
VC Timeline
2025
2026
Drag to explore the timeline
| Date | Heading | Description |
| --- | --- | --- |
| March 28th | CoreWeave IPOs at $40/share, | the largest US tech IPO since 2021, raising $1.5B and validating the GPU-as-a-service model for AI |
| June 10th | Meta buys 49% of Scale AI for $14.3B, | hiring founder Alexandr Wang to lead a new superintelligence team |
| February 2nd | Waymo raises $16B at $126B valuation, | the largest autonomous vehicle funding round ever, led by Alphabet alongside Sequoia |
| February 3rd–5th | The SaaSpocalypse: $285B erased from SaaS stocks, | triggered by Claude Cowork's launch showing AI agents replacing entire software workflows |
| February 22nd | Citrini Research publishes "2028 Global Intelligence Crisis," | a viral 97-page report predicting 40% of SaaS companies will be obsolete by 2028 |
| Q1 2026 | Global VC hits $300B in a single quarter, | with AI accounting for 80% of all venture funding. The most concentrated capital allocation in VC history |
| May 11th | OpenAI launches DeployCo, | a new enterprise deployment company with BBVA, Goldman Sachs, SoftBank, TPG, McKinsey, and Bain |
| May 14th | Cerebras IPOs, | the first AI-native chipmaker to go public at a $95B market cap |
| May 2026 | Michael Burry shorts semiconductors and AI, | telling 200K+ Substack subscribers "the market has jumped the shark" |
| June 12th | SpaceX IPOs. | The largest IPO in history, raising $75B at a $1.77T valuation, and closing day one at $2.1T market cap |
| June 16th | Cursor acquired for $60B, | the largest software acquisition ever, bringing the fastest-growing developer tool under the SpaceX-xAI |
Source: CoreWeave, Meta, Waymo, Citrini Research, OpenAI, Cerebras, SpaceX, Anysphere, PitchBook, CNBC, Bloomberg
69
## Capital Inflection
The capital inflection point for AI
Global venture capital funding to AI companies reached a record $242B in Q1 2026
| Value | Label line 1 | Label line 2 |
| --- | --- | --- |
| 18 | Q1 | 2023 |
| 14 | Q2 | |
| 17 | Q3 | |
| 16 | Q4 | |
| 16 | Q1 | 2024 |
| 27 | Q2 | |
| 25 | Q3 | |
| 50 | Q4 | |
| 66 | Q1 | 2025 |
| 44 | Q2 | |
| 49 | Q3 | |
| 56 | Q4 | |
| 242 | Q1 | 2026 |
After three years of steady quarterly increases, Q1 2026 marked a structural break: **$242B deployed in a single quarter, more than the combined total of 2023 ($65B).**
The acceleration reflects a convergence of mega-rounds in foundation model companies, surging infrastructure spend, and the **emergence of agentic startups attracting early-stage capital** at unprecedented scale.
For Latin American founders, this wave signals that the global capital base is **firmly committed to AI as an asset class.**
Sources: Crunchbase Global Venture Funding Report, Q1 2026
75
© Hi Ventures 2026 / hi.vc June 2026
## Expert Insight
REGIONAL OUTLOOK
Can AI compress development?
insights from Federico Antoni
“
For the first time in decades, **technology may allow development to move faster than history.**
Federico Antoni
MANAGING PARTNER, HI VENTURES
Source: Hi Ventures interview with Federico Antoni
| Heading | Body |
| --- | --- |
| Latin America's AI moment | Latin America may be better positioned for the AI era than many observers assume. The region combines industrial scale, entrepreneurial talent, and a strong track record of technology adoption: Brazil is one of the world's largest ChatGPT markets, Mexico shows unusually high openness to AI, and Argentina has emerged as a leader in crypto innovation. **Because AI is being adopted faster than previous technological revolutions, it creates a rare opportunity to accelerate productivity growth, strengthen competitiveness, and attract investment.** |
| Productivity, competitiveness, and investment | For decades, Latin America has struggled to close productivity gaps with more developed economies, limiting competitiveness and reducing its ability to attract investment at scale. AI may offer a different path. As a general-purpose technology capable of enhancing both knowledge work and physical production, **AI has the potential to increase productivity across entire sectors simultaneously and accelerate development timelines that historically took decades to achieve.** |
© Hi Ventures 2026 / hi.vc June 2026
## Blank slide
Can AI compress development?
insights from Federico Antoni
Four Lenses of AI Adoption in Developing Economies
Source: Four Lenses of AI Adoption in Developing Economies, Antoni, Ciesinski, Bhatia, Stanford, 2025
Source: Hi Ventures interview with Federico Antoni
THE FLYWHEEL BEHIND AI ADOPTION
What makes this moment different from previous technology waves is a reinforcing flywheel at its core. Greater adoption generates more data and workflow integration, improving products and increasing economic value. As costs fall and performance improves, adoption accelerates further. AI agents may amplify this dynamic by moving beyond assistance and increasingly performing tasks on behalf of users and organizations.
THREE PATHWAYS TO ACCELERATED DEVELOPMENT
| Text |
| --- |
| **Services and productivity.** Financial services, healthcare, education, and customer support may be among the first sectors where AI delivers productivity gains, expands access, and lowers costs. |
| **Physical AI and competitiveness.** Robotics and automation could strengthen competitiveness across manufacturing, logistics, agriculture, and mining. |
| **Infrastructure and investment.** AI is driving demand for data centers, energy, connectivity, and digital infrastructure, creating new opportunities to attract investment. |
Realizing this opportunity will require coordinated action across the public and private sectors, including support for entrepreneurship, AI talent development, regulatory experimentation, and infrastructure deployment.
72
## Blank slide
VC firms:
key insights
| Icon | Body |
| --- | --- |
| :icon-search-check: | **91% of VCs use AI to analyze investment opportunities**, up from 71% in 2025 and 45% in 2024 |
| :icon-cpu: | **68% of VCs have more than 50% of their portfolio actively deploying AI **(vs. 51% in 2025 and 52% in 2024) |
| :icon-trending-up: | 61% of VCs report that **over 60% of new investments in the last 12 months had AI embedded **in the core product. |
| :icon-brain-circuit: | 57% of VCs say that **over 25% of their portfolio is deploying AI agents specifically**, with 33% reporting more than half. |
## Blank slide
Most VCs say that over half of their portfolio companies are already deploying AI
What percentage of your portfolio companies is actively deploying AI?
Source: Hi Ventures - State of AI 2024, 2025 and 2026; n=33 (2024), n=42 (2025), n=54 (2026) venture capital firms
78 © Hi Ventures 2026 / hi.vc June 2026
| Label | 2024 | 2025 | 2026 |
| --- | --- | --- | --- |
| Less than 25% | 25 | 17 | 15 |
| 25-50% | 17 | 23 | 32 |
| 50-75% | 26 | 29 | 30 |
| Over 75% | 23 | 21 | 42 |
## Blank slide
Claude dominates among apps penetrations in VC, but few push the frontier as proprietary models are rare
Which AI applications are you using internally at your firm?
**Claude leads** penetration, overtaking ChatGPT as the most used AI tool among VC firms. **Google Gemini enters at 65%**, and only **2%** of VCs report not using any AI tools.
Source: Hi Ventures - State of AI 2026; n=54 venture capital firms
79 © Hi Ventures 2026 / hi.vc June 2026
| Label | Value |
| --- | --- |
| Anthropic Claude | 94 |
| ChatGPT / OpenAI | 69 |
| Google Gemini | 65 |
| Pitchbook / Crunchbase AI | 26 |
| Notion AI | 24 |
| Affinity / Harmonic | 17 |
| Perplexity AI | 15 |
| Microsoft Copilot | 7 |
| GitHub Copilot / Cursor | 6 |
| Other | 7 |
| Don't use AI tools | 2 |
## Blank slide
For most VCs, AI in the core product has become key for new investments
What percentage of your new investments in the last 12 months have AI embedded in the core product?
61%
of VCs report that **over 60% of new investments in the last 12 months had AI embedded in the core product**
28%
say that **more than 80% of their new investments** had AI embedded.
Source: Hi Ventures - State of AI 2026; n=54 (2026) venture capital firms
80 © Hi Ventures 2026 / hi.vc June 2026
| Label | Value |
| --- | --- |
| Less than 20% | 11 |
| 20-40% | 17 |
| 40-60% | 11 |
| 60-80% | 33 |
| More than 80% | 28 |
## Blank slide
Majority of VCs now use AI to analyze opportunities
What do you use AI for internally? (% of VC respondents)
Source: Hi Ventures - State of AI 2023, 2024, 2025 and 2026; n=33 (2024), n=42 (2025), n=54 (2026) venture capital firms
81 © Hi Ventures 2026 / hi.vc June 2026
| Label | 2023 | 2024 | 2025 | 2026 |
| --- | --- | --- | --- | --- |
| Fund Operations | 8 | 36 | 31 | 54 |
| LP Reporting | 22 | 52 | 38 | 52 |
| Due Diligence & Deal Analysis | 33 | 45 | 71 | 91 |
| Market Research | 38 | 73 | 81 | 85 |
## Blank slide
Market research is the top internal AI use case for VCs
Beyond deal sourcing and analysis, how is your fund using AI internally?
Multiple choice question
Market research leads at **85%**, but the real story is depth: **six use cases exceed 50%** adoption.
**LP reporting** and **legal review** both hit **52%** — showing AI has moved well beyond research into core fund operations.
Only **4%** say they don't use AI beyond sourcing — internal adoption is near-universal.
Source: Hi Ventures - State of AI 2026; n=54 venture capital firms (multi-choice)
82 © Hi Ventures 2026 / hi.vc June 2026
| Label | Value | Highlight |
| --- | --- | --- |
| Market research & competitive intelligence | 85 | Yes |
| Portfolio monitoring & reporting | 63 | No |
| Financial modeling & valuation | 59 | No |
| Internal communications & admin | 54 | No |
| LP reporting & fundraising materials | 52 | No |
| Legal & compliance review | 52 | No |
| Other | 7 | No |
| We don't use AI beyond sourcing/analysis | 4 | No |
## Blank slide
Agents are reaching portfolio-wide scale
What percentage of your portfolio companies are deploying AI agents specifically?
33%
say that more than half of their portfolio is already deploying AI agents.
15%
don't know the extent of agent deployment in their portfolio, exposing an opportunity for adoption.
Source: Hi Ventures - State of AI 2026; n=54 (2026) venture capital firms
83 © Hi Ventures 2026 / hi.vc June 2026
| Label | Value | Highlight |
| --- | --- | --- |
| Less than 10% | 11 | No |
| 10-25% | 17 | No |
| 25-50% | 24 | No |
| More than 50% | 33 | No |
| Don't know | 15 | Yes |
## Interlude
“For decades, Latin American founders learned to build through complexity. In the age of AI agents, that experience may become one of the region's greatest advantages.”
Jimena Pardo
Managing Partner, Hi Ventures
## Expert Insight
SPORTS OUTLOOK
The first AI-enabled World Cup:
insights from Marion Reimers
“
AI can tell you what happened on the field.** Humans can help explain why it mattered**
Marion Reimers
SPORTS ANCHOR
Source: Hi Ventures interview with Federico Antoni
| Heading | Body |
| --- | --- |
| AI is now inside the match and not around it | Across Mexico, the US and Canada, the 2026 World Cup is the first one where AI sits inside the game and not just broadcasting around it. Semi-automated offside calls land in milliseconds, a sensor-equipped ball that tracks every touch, and a real-time 3D model that rebuilds each play as it happens are only some of the new features we will soon take for granted. **The technology that reads the game has finally caught up with the game itself.** |
| Access to data is becoming the new competitive edge | For decades, the most sophisticated match analytics lived inside a handful of elite European clubs. **FIFA's Football AI Pro now gives all 48 national teams access to advanced match analysis,** and is the first time that kind of intelligence is available across all federations. What this brings is that these tools are now within reach for Latin America too, and that could shift how a whole region watches and consumes the game. |
| Where journalism and AI play better together | AI now measures the game and even tells parts of the story in real time, faster than any newsroom could on its own. That frees journalists to do what only they can, placing a result inside its culture and explaining why a match matters to a generation. **Together they cover the game in a way that either side could alone, and that synergy is where the most interesting work still lies ahead.** |
© Hi Ventures 2026 / hi.vc June 2026
## Company Profile
**Trinio** is an AI-native commerce orchestration platform founded in Brazil. It replaces shelf solutions with a system that adapts to each retailer's process, covering checkout, order management, and fulfillment routing.
AI-native commerce:
Trinio lifted checkout conversion by 12% and cut order cancellations by over 20% using AI agents to orchestrate the full order cycle
| Index | Body |
| --- | --- |
| 01 | Enterprise retailers in Latin America operate across **hundreds of stores and distribution centers**. Their systems are shelf products that force companies to work around the tool instead of the other way around.
Inventory gets misrouted and orders ship from the wrong location. **The complexity of omnichannel fulfillment at scale outgrows any static set of rules.**
|
| 02 | Trinio built an AI-native orchestration layer that sits between checkout and fulfillment:
●**Logistics decisions in real time.**
●**Flags anomalies and auto-applies new rules** without code.
●Customize flows in **natural language.**
The platform runs on a data flywheel where AI agents progressively replace static rules and learn from every order to improve routing, cost, and delivery time.
Cashea's proprietary **risk models learn from first-party repayment data** generated directly on its platform.
|
| 03 | Checkout **conversion increased by 12% on average**, adding revenue without additional acquisition spend.
Order cancellations reduced by **more than 20%** both customer-side and merchant-side.
Many other custom business rules **applied in real time** during the order lifecycle unleashing unmatched personalized workflows.
|
65
© Hi Ventures 2026 / hi.vc June 2026
## Blank slide
**Robot.com** is a Colombian-born robotics company that designs, builds, and deploys autonomous robots for industrial environments. The company is transitioning from delivery robots to humanoid for warehouses and manufacturing.
Humanoid robotics:
from delivery robots to industrial humanoids, Robot.com has completed 2.5 million tasks and is now scaling across 30 US states
87 © Hi Ventures 2026 / hi.vc June 2026
| Heading | Body |
| --- | --- |
| WHY INDUSTRIAL ROBOTICS IS READY NOW | The largest employers in logistics, food service, and manufacturing are reaching a structural inflection point. Positions go unfilled and **the cost of not automating now exceeds the cost of automating**. At the same time, advances in AI, sensors, and hardware are making it possible to build robots that are practical enough for a factory floor and affordable enough to justify at unit level. |
| LATIN AMERICA AS ROBOTICS INFRASTRUCTURE | Latin America is historically known for long working hours and low productivity. Robotics changes that equation by introducing predictability into systems that have always depended on variable human performance. **The real competitive advantage will come from generating proprietary training data faster than anyone else.** |
## Interlude
“It only takes two facing mirrors to build a labyrinth.”
Jorge Luis Borges
Argentinian writer
## Outlook
Appendix
06
## Blank slide
Survey demographics
Survey demographics
Company breakdown
420
Type of company
| Label | Value | Color |
| --- | --- | --- |
| VC Fund | 54 | var(--color-3) |
| Corporate | 98 | var(--color-2) |
| Startup | 268 | var(--color-1) |
Startups
Industry breakdown
268
Industry
| Label | Value | Color |
| --- | --- | --- |
| Others¹ | 84 | var(--color-3) |
| Media & Mktg | 12 | #ed8b00 |
| Education | 13 | #000000 |
| Retail | 15 | var(--color-4) |
| Healthtech | 21 | var(--color-5) |
| Fintech | 37 | var(--color-2) |
| B2B SaaS | 86 | var(--color-1) |
Corporates
Industry breakdown
98
Industry
| Label | Value | Color |
| --- | --- | --- |
| Others¹ | 57 | var(--color-3) |
| Retail | 19 | var(--color-2) |
| Fin. services | 22 | var(--color-1) |
VCs
AUM breakdown
54
Total AUM
| Label | Value | Color |
| --- | --- | --- |
| > 500M | 9 | var(--color-4) |
| 250-500M | 7 | var(--color-5) |
| 100-250M | 9 | var(--color-3) |
| 50-100M | 13 | var(--color-2) |
| < 50M | 16 | var(--color-1) |
Sources: Hi Ventures - State of AI 2026 survey conducted from May to June 2026
91
© Hi Ventures 2026 / hi.vc June 2026
## AI Stack Layers
The AI stack underwent a **system-level shift** across four layers
| Index | Name | Description |
| --- | --- | --- |
| 04 | Agents | AI executes tasks end-to-end, while triggering one or more workflows, integrating across tools, and delivering outcomes autonomously. |
| 03 | Applications | AI capabilities are embedded in business processes and user journeys. AI moved from specific solutions to core workflows. |
| 02 | Models | Foundation models became more powerful lowering entry barriers, enabling a broader product integration. |
| 01 | Infrastructure | Cloud providers and chipmakers scaled infrastructure capacity dramatically. |
## Blank slide
HI AI Readiness Index: Methodology
| Label | Body |
| --- | --- |
| Summary | The Hi AI Readiness Index measures how prepared organizations are to adopt, scale, and govern artificial intelligence. It is built from survey responses and provides a composite score from 0 to 100 across five pillars: Strategy & adoption (30%), Investment intensity (20%), Infrastructure & Data (15%), Governance & Risk (15%), Talent & Culture (20%) |
| Scores computation | Each respondent is scored across the five pillars using specific survey questions
• Each question is tied to one specific pillar
• Responses are normalized to a 0–100 scale
• Scores are based on checkboxes ticked, budget brackets, and yes/no inputs
• If a pillar has missing data for a respondent (some questions were optional), this responded is excluded
• Final score is a weighted average across all pillars |
| Aggregation by country | Countries are grouped into five buckets: Chile, Argentina, Brazil, Mexico and Colombia. USA and other regions were excluded from analysis given sample-size. |
| Data scope | The index is based on 395 responses from startups, corporates, and VCs
Median score per country bucket was used to avoid outlier and sample-size bias
It is not a census and results reflect the sample, not the full ecosystem |
93
© Hi Ventures 2026 / hi.vc June 2026
## Glossary (1/3)
Glossary (1/3)
| Term | Definition |
| --- | --- |
| Agent | AI systems capable of executing tasks autonomously by reasoning, triggering tools, and learning over time. |
| Agentic systems | Architectures designed around autonomous agents that can collaborate and self-correct. |
| AI governance | Policies, frameworks, and oversight mechanisms that organizations use to develop, deploy, and manage AI systems responsibly. |
| AI-native | Companies or products designed from the ground up with AI as a core component, rather than adding AI to existing solutions. |
| AIO (Agentic interaction optimization) | A new digital paradigm where AI agents interact directly with services, replacing traditional user interfaces. |
| AI Readiness Index | A composite score measuring how prepared organizations are to adopt, scale, and govern AI across five pillars. |
| AI stack | The layered architecture of AI systems, spanning infrastructure, foundation models, applications, and autonomous agents. |
| Benchmark | A standardized test or dataset used to measure and compare AI model performance across tasks. |
| Compute | The processing power required to train and run AI models, a key cost driver in AI deployment. |
94
© Hi Ventures 2026 / hi.vc June 2026
## Glossary (2/3)
Glossary (2/3)
| Term | Definition |
| --- | --- |
| Context window | The amount of information (measured in tokens) a model can consider in a single prompt. |
| Copilot | An AI assistant that works alongside humans, suggesting actions or content without acting autonomously. |
| Data flywheel | A self-reinforcing cycle where AI-generated insights produce more data, which further improves model performance. |
| Fine-tuning | The process of customizing base AI models with specific data to improve performance. |
| Foundation models | Large-scale AI models trained on broad data to perform a wide range of tasks, often serving as the base for fine-tuning or specialized applications. |
| Function calling | Allows AI models to execute structured actions (like calling APIs) within a workflow. |
| H100 | NVIDIA's high-performance chip used to train and run large AI models. |
| Hallucination | When an AI model generates outputs that appear plausible but are factually incorrect or fabricated, cited as the top deployment risk. |
| Human-in-the-loop | An operational model where humans review and approve AI agent actions before execution, the dominant autonomy approach in current deployments. |
95
© Hi Ventures 2026 / hi.vc June 2026
## Glossary (3/3)
Glossary (3/3)
| Term | Definition |
| --- | --- |
| Inference | The process of generating outputs from a trained AI model. |
| LLM (Large language model) | AI models trained on large text datasets to perform reasoning, generation, and comprehension tasks. |
| Memory | The ability of AI agents to retain information across sessions. |
| Multimodal models | AI systems that understand and generate different data types, such as text and images. |
| Nearshoring | Relocating business operations to nearby countries, increasingly relevant as AI-driven automation reshapes manufacturing. |
| Open-source model | AI models with publicly available architecture and weights. |
| Orchestration | The coordination of agents, tools, and workflows to automate complex tasks. |
| Prompt engineering | The practice of crafting model inputs to guide output behavior. |
| Reasoning models | AI models designed to break down complex problems step by step before generating an answer, improving accuracy on analytical tasks. |
96
© Hi Ventures 2026 / hi.vc June 2026
## Hi Flappy
Thank you :)
let's play!
Click or tap to start playing
Score
Nice try!
Add your score to the board
Your name
Save
Leaderboard
Tap or press space to play again!