Apoorva Mehta built Instacart from a grocery delivery idea into a company worth billions. Now he’s trying to build a hedge fund where human portfolio managers are optional.

Mehta launched Abundance in 2025 with a small team of quantitative researchers, engineers, and AI specialists. In April 2026, he closed $100 million in seed funding and went public with the thesis: AI agents (thousands of them) will run every meaningful function in the fund.

That means AI identifies trading ideas, conducts the research, picks stocks for long and short positions, sizes the bets, and executes the trades. The human team builds and maintains the models. The models do the rest.

Mehta has said Abundance’s returns have outperformed multiple market indexes, though he’s declined to name which benchmarks. That’s a detail investors will press on over time, but the structure itself is what’s worth paying attention to.

The Shift That Triggered This

Mehta has pointed to OpenAI’s o3 model as the catalyst. When o3 demonstrated it could reason through genuinely complex, multi-step problems (not just pattern-match on training data), Mehta saw a different kind of AI capability emerging.

Most hedge funds have been using AI for years as a support tool: screens, backtests, sentiment analysis fed to human portfolio managers who make the final calls. Abundance is designed differently. The AI isn’t advising. It’s deciding.

The distinction matters. A lot of “AI-powered” business tools are really AI-assisted tools dressed up in better marketing. What Abundance represents is an attempt to build a fully autonomous AI workforce for a cognitively demanding domain, one where decisions have direct financial consequences in real time.

Not Just a Finance Story

Finance journalists are covering this because it’s a hedge fund. Business leaders should be paying attention because it’s a workforce model.

The question Abundance is testing is not “can AI help with financial research?” That’s been answered. The real question is: at what point can you replace the core decision-making function of a knowledge worker (not just the administrative wrap around it) with an AI agent that runs continuously, improves over time, and doesn’t need a salary, a performance review, or a break?

That question is being tested right now in finance. The answer will be instructive for every other knowledge-work function in business.

The jobs at risk aren’t what most people think

The first wave of automation covered the obvious stuff: data entry, scheduling, document processing. Abundance represents the second wave, where AI runs judgment-intensive work at scale.

Portfolio research is fundamentally about gathering information from multiple sources, synthesising it against a framework, making a probabilistic judgment under uncertainty, and acting on it. That description also fits: business intelligence analysis, customer success planning, legal due diligence, strategic planning, and procurement decisions.

If AI agents can do it with stocks, the architecture works for a lot more than stocks.

What Abundance Is Actually Building

The fund employs thousands of AI agents working in parallel. Not one AI assistant, but a distributed workforce of agents, each handling specific parts of the research and decision-making pipeline. The human team’s job is to build the systems, evaluate the outputs, and evolve the models.

This is closer to the “AI employee” model than the “AI tool” model. The agents aren’t producing reports for humans to act on. They’re producing actions, with humans overseeing the quality of the system rather than the individual decisions.

That architecture (specialised agents coordinated around a common goal, with humans managing the system rather than doing the work) is exactly where the most ambitious enterprise AI deployments are heading.

What This Means for Business

Most business owners thinking about AI are still in the tool-adoption phase: new software, automation of repetitive tasks, co-pilots for writing or coding. That’s the right starting point, but Abundance signals something about where this is going.

The shift from “AI tool” to “AI workforce” is not a distant scenario. It’s being tested right now at a firm with $100 million in backing and a founder who has already built at scale once before.

A few practical implications:

Know what your humans are actually doing. The tasks most exposed to this shift are the ones that look like research plus judgment plus action. If your team spends significant time gathering data, synthesising it, and making decisions from it, that workflow is increasingly automatable, not just acceleratable.

The gap between AI-assisted and AI-run is closing faster than most governance teams realise. Abundance’s timeline from founding to $100M raise is under two years. The infrastructure for autonomous agent workflows now exists off the shelf in ways it simply didn’t three years ago.

The firms that win won’t be the ones with the most AI tools. They’ll be the ones that figured out how to give AI agents real authority over real workflows, with the right oversight model to catch failures before they compound.

The story of Abundance is partly about finance. But mostly it’s about what a real AI workforce deployment looks like when someone with the skills and capital to try it actually does.

---

Enterprise DNA helps businesses understand and deploy AI. From data literacy training through EDNA Learn to fully deployed AI agent workforces through Omni Ops. If you’re thinking through what an AI workforce model means for your organisation, we’d like to talk.