The Next 10 Ventures: Where AI Gens Is Placing Bets in 2026-2027_
The Next 10 Ventures: Where AI Gens Is Placing Bets in 2026-2027
Every venture firm has a thesis. Most theses are vague enough to be unfalsifiable: "We invest in exceptional founders building transformative technology." Translated: we invest in whatever seems good at the time.
We want to be more specific. Not because specificity guarantees returns — it doesn't — but because it forces intellectual honesty. If we tell you exactly what we're looking for, you can evaluate whether we found it. And if we're wrong about the categories, the specificity means we'll know we were wrong, rather than rewriting history to pretend we were right all along.
What follows is our thesis map for 2026-2027. Ten areas where we see venture-scale opportunity that aligns with AI Gens' capabilities — platform engineering, AI, infrastructure — and where timing signals suggest the next 18 months are critical for company formation.
Some of these we'll build internally. Some we'll invest in. Some we'll co-build with founders who bring domain expertise we lack. The distinction matters less than the conviction: these are the areas where we believe outsized value will be created.
1. Vertical AI Agents for Regulated Industries
Market size: $45B+ (healthcare compliance alone is $28B; financial regulation technology is $17B and growing at 22% CAGR)
Why now: Horizontal AI agents are commoditizing rapidly. ChatGPT, Claude, Gemini — they're all converging on similar general capabilities. But regulated industries need agents that understand specific compliance frameworks, audit requirements, and liability boundaries. A healthcare compliance agent needs to know HIPAA, HITECH, state-specific regulations, and payer-specific rules. A fintech regulation agent needs to understand BSA/AML, KYC requirements across jurisdictions, and real-time sanctions screening. This domain knowledge creates moats that horizontal agents can't easily breach.
YC's Spring 2026 RFS explicitly calls out "AI for industries that require licensing" — recognizing that regulated domains are where AI creates the most value precisely because the cost of human compliance expertise is highest.
What we'd build: Agent systems that combine LLM reasoning with structured regulatory knowledge graphs. Not chatbots that answer compliance questions — agents that actively monitor, flag, and remediate compliance issues in real-time. The agent doesn't replace the compliance officer. It gives the compliance officer superhuman coverage.
Founder profile: Former compliance officers or regulatory technologists who've felt the pain firsthand. Deep domain expertise is non-negotiable — you can teach someone to build AI, but you can't teach someone twenty years of regulatory intuition in a crash course.
2. WhatsApp-Native Operator OS for LATAM SMBs
Market size: $40B+ (LATAM SMB software market, largely unaddressed by existing SaaS)
Why now: 93% of Brazilian SMBs use WhatsApp as their primary business communication tool. Not as a supplement to email or Slack — as the entire operating system for their business. Orders come in via WhatsApp. Invoices go out via WhatsApp. Inventory is tracked in WhatsApp groups. Employee scheduling happens in WhatsApp.
Yet virtually no software company has built a full operator OS native to WhatsApp. The existing SMB SaaS stack — CRMs, ERPs, inventory management — assumes a web browser or mobile app as the primary interface. For a bakery owner in Sao Paulo or a mechanic in Medellin, that assumption is wrong. Their interface is WhatsApp. Build for that reality instead of against it.
Meta's expansion of WhatsApp Business API and the introduction of WhatsApp Flows make this technically feasible at scale for the first time.
What we'd build: A complete business operating system — orders, payments, inventory, scheduling, customer management — that lives entirely within WhatsApp. No separate app to download. No web dashboard to learn. The SMB owner manages their business in the same interface where they talk to their family.
Founder profile: Someone who's run or deeply served LATAM SMBs and understands the operational reality. Ideally bilingual (Portuguese/Spanish and English), with the empathy to build for users who've never used traditional enterprise software.
3. AI Infrastructure Governance
Market size: $12B by 2028 (projected from current AI infrastructure spend of $200B+ and typical governance/security spend ratios)
Why now: As enterprises deploy agentic AI systems, they face a governance vacuum. Who audits what the agent did? Who sets policy for what it's allowed to do? Who controls costs when agents can autonomously provision resources? The security, compliance, and cost-control frameworks that exist for human workers and traditional software don't map to autonomous agents.
a16z's Big Ideas 2026 list includes "AI governance infrastructure" as a category — recognizing that the explosion of agent deployments creates urgent demand for guardrails, audit trails, and policy frameworks.
What we'd build: A governance layer that sits between AI agents and the systems they interact with. Policy definition (what agents can do). Audit logging (what agents did do). Cost controls (what agents are allowed to spend). Anomaly detection (when agents behave unexpectedly). Think of it as the operating system for managing a fleet of AI workers.
This is directly adjacent to what we're building with Apollo — our delivery control plane already governs 203 agent tools. The question is whether that governance framework generalizes into a standalone product.
Founder profile: Infrastructure engineers with security or compliance backgrounds. People who've built IAM systems, policy engines, or audit infrastructure and see the parallel to governing autonomous agents.
4. Hospitality Tech (HOST360 Expansion)
Market size: $30B (hospitality technology, growing at 12% CAGR post-COVID recovery)
Why now: The hospitality industry is simultaneously experiencing a labor shortage and a technology adoption wave. Properties that relied on large staff are being forced to automate. But the technology they're adopting — property management systems, channel managers, guest experience platforms — is fragmented, disconnected, and designed for a single-property paradigm.
Multi-property operators — the fastest-growing segment — need unified platforms that span brands, locations, and channels. HOST360, our incubated venture in this space, has validated the thesis. Now we're looking to expand: either by accelerating HOST360's growth or by investing in complementary ventures that fill gaps in the hospitality tech stack.
What we'd build/invest in: Unified property operations platforms for multi-brand operators. AI-powered dynamic pricing across channels. Guest experience orchestration that spans pre-stay, in-stay, and post-stay. Maintenance prediction and workforce optimization. Any technology that helps a hospitality operator run more properties with fewer manual processes.
Founder profile: Hospitality operators who became technologists, or technologists who spent years embedded in hospitality operations. The industry has deep domain knowledge requirements — rate management alone has nuances that take years to learn.
5. Consumer Brand Engineering with AI
Market size: $22B (brand management and creative technology)
Why now: Building a consumer brand used to require armies of creative professionals, months of agency work, and millions in production costs. AI has compressed this dramatically. A small team can now generate brand assets, test positioning, produce content, and iterate on creative direction at a pace that was impossible two years ago.
But the tools are fragmented. Midjourney for images. ChatGPT for copy. Eleven Labs for voice. RunwayML for video. A brand team stitches these together manually, and brand consistency suffers. There's no unified platform for AI-native brand engineering — one that maintains brand guidelines, enforces consistency, and orchestrates multiple generative AI tools under a single creative direction.
This connects directly to Mustard, our incubated brand engineering venture.
What we'd build: An AI-native brand engineering platform that ingests brand guidelines and produces consistent creative assets across channels, formats, and languages. Not individual AI tools — an integrated system that ensures every asset is on-brand, on-voice, and on-strategy. The platform acts as a brand guardian powered by AI.
Founder profile: Creative directors who code, or engineers who've led brand teams. The intersection of brand intuition and technical architecture is rare and valuable.
6. Cross-Border Fintech on the Brazil Stack
Market size: $35B (cross-border payments in LATAM, growing at 25% CAGR)
Why now: Brazil has assembled the most advanced real-time financial infrastructure in the world. Pix processes over 200 million transactions per day. Open Finance has 30+ million consented users sharing banking data. Drex — Brazil's CBDC — is entering pilot with programmable money capabilities. Together, they form the "Brazil Stack": a foundation for financial products that simply aren't possible in markets without comparable infrastructure.
The opportunity is cross-border. Brazilian SMBs export $80B+ annually, mostly to the US, EU, and other LATAM countries. The existing cross-border payment rails are slow (3-5 days), expensive (3-7% fees), and opaque. The Brazil Stack enables real-time, low-cost, transparent cross-border transactions — but nobody has built the user-facing products that leverage this infrastructure for SMB exporters.
Sequoia's $950M seed and venture funds signal that fintech infrastructure remains a category with massive capital behind it, and LATAM is specifically called out as an underserved market.
What we'd build: Cross-border payment and treasury management products built natively on Pix, Open Finance, and Drex. Instant settlement for SMB exporters. Automated FX hedging using programmable money. Working capital products underwritten by real-time cash flow data from Open Finance. The full stack from payment initiation to settlement, without legacy banking rails.
Founder profile: Fintech builders who understand both the Brazil Stack's technical capabilities and the operational reality of cross-border commerce. Ideally someone who's built on Pix/Open Finance APIs and has experience in trade finance or treasury management.
7. MLOps/LLMOps Productization
Market size: $18B (MLOps market, growing at 38% CAGR as AI adoption accelerates)
Why now: The stat that keeps us up at night: 85% of AI pilots fail to reach production. Not because the models don't work — because the infrastructure to deploy, monitor, version, and govern models in production doesn't exist in most organizations.
The gap between a working Jupyter notebook and a production AI system is enormous. Model versioning. Data pipeline orchestration. Inference optimization. Drift detection. Cost management. Compliance logging. A/B testing for model variants. Each of these is a solved problem in isolation, but the integrated platform that handles all of them — purpose-built for LLMs and generative AI — doesn't exist yet.
This is the prototype-to-production gap we've written about extensively. It's not a technology problem. It's an infrastructure problem. And infrastructure is what we build.
What we'd build: An integrated LLMOps platform that takes a model from prototype to production with opinionated defaults. Not another experiment tracking tool or model registry — a deployment-first platform that handles the entire lifecycle: containerization, serving, scaling, monitoring, versioning, cost tracking, and compliance. Think Vercel for AI models: push your model, it's in production, with guardrails built in.
Founder profile: ML engineers who've felt the production pain — people who've spent months bridging the gap between a working model and a reliable production system. Strong infrastructure engineering skills are essential; ML research skills are secondary.
8. Agentic PSA Platforms
Market size: $16B (professional services automation, but the category is being redefined by AI)
Why now: Professional services firms — consultancies, agencies, managed service providers — run on PSA (Professional Services Automation) platforms. These platforms handle project management, resource allocation, time tracking, invoicing, and reporting. The incumbents — ConnectWise, Autotask, Mavenlink — are built for a world where humans do the work and track time against projects.
That world is ending. When AI agents handle 40-60% of the work in a professional services engagement, the PSA platform needs to track agent activity alongside human activity. Task assignment should flow to agents or humans based on capability, not just availability. Cost tracking needs to include compute costs alongside labor costs. Billing needs to reflect outcomes, not hours.
The entire PSA category needs to be rebuilt for the agentic era: task in, code out, deploy done, cost tracked, client billed. End to end.
What we'd build: A PSA platform designed for hybrid human-agent teams. Intelligent task routing (agent vs. human). Unified cost tracking (compute + labor). Outcome-based billing engine (not time tracking). Real-time project health dashboards that reflect both human and agent throughput. Integrated deployment pipelines so that "done" means deployed, not "merged to main."
Founder profile: Former professional services leaders who've tried to make legacy PSA tools work for AI-augmented delivery and found them fundamentally broken. Deep understanding of services economics — margins, utilization, realization rates — combined with engineering ability.
9. AI-Augmented Real Estate Investment Analysis
Market size: $20B (real estate technology, with investment analysis being the highest-value segment)
Why now: Real estate investment analysis is still shockingly manual. Analysts spend weeks building financial models for a single deal. Comparable property analysis requires pulling data from multiple brokers, county records, and market reports. Environmental and zoning due diligence involves reading hundreds of pages of municipal documents. Construction cost estimation relies on outdated databases and expert intuition.
AI can compress this from weeks to hours. Not by replacing the analyst's judgment, but by automating the data gathering, standardizing the analysis framework, and flagging risks that humans miss because they can't process the volume of relevant information.
The real estate market is cyclical, and cycles create opportunity for technology adoption. The current rate environment is forcing investors to be more disciplined in their analysis — which means better tools have more leverage.
What we'd build: An AI-powered investment analysis platform that ingests property data, market comparables, zoning regulations, environmental reports, and construction cost databases to produce comprehensive investment memoranda in hours instead of weeks. The platform doesn't make the investment decision — it gives the investor superhuman analytical capability to make better decisions faster.
Founder profile: Real estate professionals who've built financial models and felt the pain of manual data gathering. Understanding of real estate capital structures (debt, equity, mezzanine) is essential. Technical co-founders with NLP/document processing experience to handle the diverse input formats.
10. Construction Intelligence / Fintech
Market size: $25B (construction technology, with fintech being the fastest-growing subsegment at 30% CAGR)
Why now: Construction is a $13 trillion global industry that still runs on spreadsheets, paper invoices, and handshake agreements. The financing layer is even worse: construction loans involve draw schedules tied to inspection milestones, with manual verification at every stage. A $5 million construction loan might require 30+ site inspections, each generating paperwork that's faxed (yes, faxed) to the lender.
This is the problem space that inspired Delfin, and our experience there has convinced us that construction intelligence — the combination of project data, financial analysis, and risk assessment — is a venture-scale opportunity.
The timing is driven by two forces: a structural housing shortage (the US alone needs 4-6 million additional housing units) that's accelerating construction activity, and a generational shift in construction companies as founders retire and successors demand modern tools.
What we'd build: Platforms that combine construction project intelligence with financial products. AI-powered draw management that verifies construction progress from satellite imagery, drone data, and IoT sensors — replacing manual inspections. Risk scoring for construction loans based on real-time project data rather than quarterly inspections. Working capital products for subcontractors who currently wait 60-90 days for payment.
Founder profile: Construction industry veterans — project managers, general contractors, or lenders — who understand the operational complexity of construction finance. The regulatory knowledge (construction liens, prevailing wage, bonding requirements) is deep and domain-specific. Technical co-founders with experience in computer vision, geospatial analysis, or financial modeling.
The Common Thread
Look across these ten areas and a pattern emerges. Every one of them involves:
- A regulated or domain-complex industry where general-purpose tools fail because they lack context.
- An infrastructure gap where the technology exists in pieces but hasn't been assembled into an integrated platform.
- A timing signal — regulatory change, technology maturation, market shift — that makes now the right moment for company formation.
- Alignment with AI Gens' capabilities in platform engineering, AI, and infrastructure building.
We're not trying to be everything to everyone. We're trying to be the best venture studio for infrastructure companies that leverage AI in complex domains. If a venture doesn't fit that description, we pass — no matter how attractive the market size looks on a slide.
A Note on Sources and Conviction
This thesis draws from several sources: YC's Spring 2026 Request for Startups, which emphasizes AI for licensed industries, developer infrastructure, and LATAM-focused fintech. a16z's Big Ideas 2026 list, which highlights AI governance, vertical SaaS transformation, and construction technology. Sequoia's recent $950M raise across seed and venture funds, signaling sustained conviction in infrastructure and fintech.
But lists and fund sizes don't create conviction. Experience does. We've built in hospitality tech (HOST360), brand engineering (Mustard), delivery infrastructure (Apollo), and construction fintech (Delfin-inspired). Our thesis isn't theoretical — it's informed by the operational reality of building companies in these spaces.
Some of these ten bets will be wrong. Markets will shift, timing will be off, execution will fall short. That's the nature of venture building. But we'd rather be specifically wrong than vaguely right. Specificity forces learning. Vagueness allows denial.
If You're Building Here
This is not an academic exercise. We're actively looking for founders building in these ten areas — or in adjacent spaces we haven't considered.
If you're a compliance officer who's building AI tools for your industry. If you're a Brazilian developer building WhatsApp-native business tools. If you're an ML engineer who's tired of watching AI pilots die in the gap between notebook and production. If you're a construction professional who knows the financing process is broken and has a plan to fix it.
We want to talk to you.
Not for a pitch meeting. For a conversation about whether AI Gens' capabilities — engineering, AI, infrastructure, capital, and the flywheel of our existing portfolio — can accelerate what you're building.
We'll bring the platform. You bring the domain. Let's build something that matters.