Financial institutions paid $3.8B in AML fines last year while spending $206B on financial crime compliance, with labor its largest cost.
Today we're launching Zomma (@ycombinator S26)
AI operations employees for finance back offices.
We start with compliance: for every cross-border transaction alert, our agent works the case end to end on the same screens your analysts use: gathers the evidence, writes the assessment, drafts the RFI. No APIs. Onboarding is an SOP and a login.
92% on OSWorld. Live compliance-ops pilots with a US RIA and a $10B AUM wealthtech.
If you run compliance or financial-crime ops, or know someone who does, let's talk → https://t.co/fJ9V0lT9ZM.
With Teach Mode, the agent asks you clarifying questions, proposes a fix to its own procedure, and gets better with every round.
Added benefit? A living procedure that stays up to date with how your company works.
If you're building agents for your operations, let's chat: https://t.co/PDIhyQQrPS
Introducing @zommalabs Teach Mode 👇
Most SOPs live in people's heads. Even when they're written down, they go out of date fast.
A good hire asks good questions, especially after making mistakes so that they learn and improve.
Does your agent do that? 👀
Are you a hedge fund deploying agents?
Backtest AI agents like you backtest strategies.
Getting an AI agent to answer one question is easy. Trusting it with management fees, across every trade, position P&L, side letter and four systems, is not.
So we test agents the way hedge funds test strategies. Your team's past decisions become the benchmark. The agent flags terms that can be read two ways, your team answers, and every change is backtested before it goes live.
In this run, an agent swarm checked every investor's fees across multiple systems. They found USD 437,390 in overcharges, each with audit and replay of every click.
Introducing Zomma Agent Fleets 🙂↕👇
It's 2026. Why is your customer ticket still stuck for days?
Because most teams still work in a single queue.
With Zomma, you can now dispatch a Fleet of parallel computer-use agents as soon as work comes in.
The biggest hack? Fleet = better outcome with fewer tokens. Each subagent gets just enough context to do the work well, saving tokens and time.
And you can score every subagent against your own eval set.
@ZommaLabs is in early access. If you're building agents for your operations, let's chat 📥
what most people still don't grasp is how many computers we'll need when agents run everything
compute isn't just GPUs for AI training or inference. It's CPUs and memory for your Muse, Instinct, Town, Claude Code, Codex sessions to always be running
this is especially true when enterprise deploys more agents (and it's still VERY early!)
everyday we dynamically scale ~500 VMs up and down for @ZommaLabs computer-use swarms for operations and it will only go up
now that agents can do real work, they just need a place to do it persistently and safely = computers on cloud
Some of my thoughts making a computer use product as a YC S26 startup founder
A bit of context - we’re building computer use agents for operations heavy work in enterprise at Zomma. Something I’ve been thinking about a lot lately is how to actually design this product. Let me be honest, enterprise deployment really sucks - they’re slow, everyone’s flow is different, SO MANY back and forths - but I think it’s a problem worth solving. Our customers can do a lot more when they don’t have to worry about just grunt work :)
So what actually matters for CUA in enterprise:
- Demos are great and all but those are still just the happy path - what happens when it fails
- Design human in the loop but do it without causing reviewers fatigue? (Really need to figure this one out)
- Passwords, 2fa, accounts?? Like sure there are ways we do this right now but it still feels forced - too much friction for the end user
- How should success be measured for these agents? Do we just put Zomma on a PIP if it doesn’t get a job done😭 - this matters a lot for pricing
- How should you interact with Zomma and at what frequency? - kind of ties in with the 2nd point
Well all of this to say I think CUA is the future (debate me in the comments). It’s probably infeasible to expect everyone in the near term to build out API’s and headless counterparts for their legacy and internal systems so why not just use it like it was designed to be used. Also if it was price people are worried about; it’s only getting cheaper…🤷♂️
Anyways go follow the company page - @ZommaLabs
After 6 weeks of building, Vercel AI Accelerator teams are ready to show their ships.
Join us in SF on April 16 for Demo Day, where founders will pitch to @Vercel, @AWSstartups, @AnthropicAI, @OpenAI, and more.
Save your seat ↓
https://t.co/yJuTaWe8rs