Y Combinator is supporting 241 generative artificial intelligence startups in 2026, reflecting both the breadth of use cases for large language models and a tilt toward industry-specific agents and workflows. The portfolio spans e-commerce, healthcare, biotech, creative industries, legal and regulatory tooling, real estate, government, and classic software infrastructure, often pairing foundation models with narrow expert domains. Many of these companies are very small, typically listing 1-7 employees, and are based in hubs such as San Francisco, New York, Boston, and Copenhagen, with a few operating fully remotely or internationally.
A large cluster of startups targets business operations and customer-facing automation through generative agents. Return Signals builds an artificial intelligence powered luxury concierge for e-commerce brands that proactively reaches customers via SMS to recommend exchanges and preserve revenue, reporting 5-10% absolute margin uplift and response rates of over 80% of eventual returns, while converting 70% of return intents into exchanges. LunaBill deploys artificial intelligence voice callers for healthcare billing teams, where insurance claim follow-up calls account for the 80% of the daily workload of a healthcare billing team, with each call averaging 30 minutes. Since its launch in July, LunaBill reports $764K in contracted ARR, including $428K in live revenue, a 10x increase in claims followed up per biller in the first week, 100% of pilots converted to paying customers, and 50,000+ automated calls. Pollen monitors email, support tickets, product usage, and CRM data so artificial intelligence agents can surface churn and upsell signals, while Paragon builds an artificial intelligence growth engine for distributors and manufacturers, and Calltree provides enterprise-grade artificial intelligence support reps that can also operate internal tools.
Another major theme is the use of generative artificial intelligence in highly regulated, scientific, or complex decision-making domains. Strand AI develops multimodal foundation models to generate missing bio-data about patients so pharmaceutical companies can select better drug trial participants. Ritivel provides artificial intelligence native workspaces for regulatory and medical writing, using agents that draft CTDs, CSRs, INDs, and BLAs in minutes instead of weeks, and notes that every month a drug is delayed from reaching the market costs pharma companies ~$45M in lost revenue. Anto Biosciences focuses on the gut microbiome with a foundation model for microbial communities to predict drug toxicity and efficacy across populations. AutoSitu embeds coordinated artificial intelligence agents inside cities’ cross-department development review workflows to catch planning issues in minutes, and Trava builds agents for global trade compliance to audit customs entries and automate classification for importers. Permitify helps city building departments 10x their productivity in plan review, positioning itself as a way to sustain profitable government branches under staffing pressure.
Developer tools, reliability infrastructure, and performance evaluation for artificial intelligence systems also feature heavily. Sonarly focuses on the run time phase of software by triaging production alerts so teams can trust what pages them, while ZeroEval provides calibrated large language model judges and automatic evaluations so companies can optimize their agents 10x faster than manual experimentation. Confident AI offers benchmarking, safeguards, and improvement tools for large language model applications using evaluation algorithms from the DeepEval open source project, which it notes has 12.6k stars and >3m monthly downloads. TrainLoop lets developers boost large language model performance using reinforcement learning, Kashikoi simulates environments to benchmark artificial intelligence agents via CPU friendly world models, and The LLM Data Company trains models for data scarce verticals. Truffle AI exposes agents as simple APIs for developers to embed into applications, assistant-ui builds an open source Typescript and React library with a backend as a service for adding chat interfaces, and Sennu AI reviews every pull request in Salesforce codebases like a senior architect would.
Consumer and creative applications form another visible pillar, often blending entertainment with generative media tooling. Martini, Koyal, Prism, Bezel, Absurd, and Pixley AI focus on generative video, film, and advertising, from professional filmmaking workflows and multilingual script-to-audio-to-video pipelines to virtual fashion shoots and short-form content creation. Pixley AI reports that in just 3 weeks after launch, over 1000 families across 75 countries began using its platform to transform kids’ drawings into animated cartoons, and outlines a vision of an “artificial intelligence Netflix” where shows are generated around each viewer’s interests and habits. Koyal cites paid pilots with major music labels and production houses, including a completed pilot with Universal Music and T-Series that produced 5 music videos for Grammy and Oscar winning artists, each crossing 1.5 million views. Dollyglot offers real time video avatars akin to Character.ai so users can converse with characters and celebrities, and Awen enables creatives to describe visual scenes in natural language and have artificial intelligence interpret and animate them using voice-driven reasoning.
Several startups are focused on search, discovery, and visibility in an emerging ecosystem of artificial intelligence agents and conversational interfaces. sitefire analyzes which content influences answers on ChatGPT, Gemini, and similar systems, then synthesizes artificial intelligence optimized, ready-to-publish pages. Bear captures and converts high-intent leads coming from artificial intelligence agents like ChatGPT and Perplexity, while Wildcard provides SKU-level visibility for e-commerce brands inside artificial intelligence shopping experiences and plans to activate next-generation agentic commerce. Operand works with retailers, manufacturers, distributors, and capital allocators to apply artificial intelligence systems to pricing, forecasting, allocation, and diligence. On the education side, Alice.tech personalizes exam prep using artificial intelligence and reports more than 10k college students using its product, $20k MRR with +10% week-over-week growth over the last 8 weeks, and a core market opportunity of +$10b within a global education market it describes as almost $10trillion.
Some of the most mature companies in the portfolio illustrate how fast generative artificial intelligence businesses can scale. Emergent builds autonomous coding agents that generate, test, and deploy production software directly from natural language, reporting that since its public launch Emergent has reached $50M ARR in 7 months, with 5M+ users across 190+ countries having built 6M+ applications, and $100M raised from a roster of investors including Khosla Ventures, SoftBank, Google, Lightspeed India, Prosus Ventures, Together Fund, and Y Combinator. Absurd highlights creative output by noting its video for Kalshi reached 1M+ views and that its videos average 400K+ organic views, while companies like Replit, Brex, and others use its artificial intelligence produced ads. Across this cohort, Y Combinator is betting on a mix of infrastructure, vertical agents, and novel consumer products that treat generative artificial intelligence as both a co-pilot and a core execution engine for digital work.
