A bill died in Nevada, and a company was born. The founders of Fed10 were paying consultants to keep one of their bills alive when a last-minute committee change slipped past everyone. By the time anyone noticed, the bill was dead. That single miss - the kind that happens in a marble hallway on a Tuesday afternoon - is the reason three people left three universities to build an AI that never looks away.
Fed10 is a legislative consulting firm, except most of the consultants are software. Its AI agents monitor every bill introduced across all 50 U.S. state legislatures and Congress. When a bill lands, the system decides whether it is a threat or an opportunity for a specific client, drafts the amendment language to fix it, and then maps out exactly who to call and what to say. The company's own summary is blunt: the same work a $500-an-hour policy consultant does, but in seconds instead of days.
The FoundersLobbyists before they could vote
The three co-founders - Armand Iorgulescu, Zayan Islam, and Winston Wei - did not read about how legislation works. They ran campaigns. As high schoolers in Nevada, they pushed a bill, AB367, through the state legislature; by their account it affects roughly 344,000 people a year. Iorgulescu became one of the youngest gubernatorial appointees in the state's history. Wei ran a government affairs desk for the state Democratic party. Islam was writing code and learning committee procedure at the same age most people are learning to drive.
Then they left. Harvard, Williams, and Berkeley, respectively. The pitch they tell is that the job they used to do by hand - reading bills, spotting the threat, writing the fix, working the phones - is a job a machine can now do faster and without sleeping through a committee vote.
It is a specific kind of founder story. Plenty of people build tools for an industry they have only read about. Fewer build the tool they personally needed and could not buy. The founders were not customers of the lobbying industry in the abstract; they were on the inside of it, watching money change hands for coverage that still left gaps. That vantage point shapes what Fed10 is and is not. It is not a general-purpose "AI for policy" wrapper. It is narrow on purpose: watch the bills, catch the ones that matter, produce the exact artifact a client needs to act.
You can't call a phone number you never saw
Here is the uncomfortable math of government affairs. There are 50 state legislatures plus Congress. In a busy year they collectively introduce well over 100,000 bills. Any one of them can quietly redefine what your business is allowed to do. No human policy team, however well paid, can read all of them, let alone read them fast enough to act before a vote.
So companies do what they can afford. The large ones hire firms and pay retainers that can run past $30,000 a month. The rest buy tracking software that tells them a bill exists but not what to do about it. Both groups, Fed10 argues, share the same failure mode: they get blindsided anyway. Spending more money has not solved a coverage problem.
The timing is what makes the whole thing fragile. Legislation does not move on a predictable schedule. A bill can sit dormant for weeks, then get amended and voted out of committee inside 48 hours. The window to influence it - to propose an edit, to reach the right staffer - is often measured in a single day. That is the window where the Nevada bill was lost. A human team covering dozens of clients across dozens of jurisdictions is, by definition, not watching all of them at 4 p.m. on the Tuesday that matters. Software does not have that constraint.
Monitor, classify, draft, advocate
Under the hood, Fed10 runs a pipeline that mirrors the four things a good lobbyist actually does in a week - compressed into one loop that runs continuously.
That last step is the tell. A tracking tool stops at "a bill was introduced." Fed10's bet is that the valuable part was never the alert - it was knowing, at 2 a.m. before a morning markup, which amendment to propose and which staffer to send it to.
Classification is where the difficulty hides. A bill about "data privacy" might be irrelevant to a logistics company and existential to a fintech one. The same words carry different stakes depending on who is reading them. Fed10's job is to hold a model of each client's business - what it does, where it operates, what could reclassify it - and read every incoming bill against that model. Get it too sensitive and you drown the client in false alarms until they stop reading. Get it too loose and you miss the one that costs $10 million. The founders' claim is that the case study proves the point: they caught a bill that carried exactly that price tag and turned it into a plan while a traditional consultant would still have been reading.
The CustomersBuilt for the companies the incumbents priced out
Fed10 is aimed at corporate policy, government affairs, and compliance teams - and specifically at the mid-sized end of the market. Think fintech, gig economy, and healthtech: industries where a single state reclassification can break the business model, but where a six-figure lobbying retainer is impossible to justify. These are companies with real regulatory exposure and no army of consultants to cover it.
Where Fed10 sits on the map
The legislative intelligence market already exists. Bloomberg Government, LexisNexis, Quorum, and FiscalNote all sell bill tracking and policy data. Traditional lobbying firms sell relationships and judgment. Fed10 is trying to occupy the seam between them: the AI depth of a software product with the "here is your move" output of a consultant.
| Approach | What you get | The gap |
|---|---|---|
| Tracking software | A feed of bills | No fix, no strategy |
| Lobbying retainer | Strategy and access | Expensive, human-limited coverage |
| Fed10 | Threat call + amendment + who to phone | Young, unproven at scale |
Software margins on a consultant's job
Fed10 sells subscription access to its platform - B2B SaaS, positioned explicitly against the $30,000-a-month consultant and the legacy tracker. The company is small: a team of five, part of Y Combinator's Winter 2026 batch, working with YC partner Jared Friedman. It has raised a $130,000 seed round tied to the batch. These are early numbers, and Fed10 does not hide it. What it is selling is not scale yet; it is a wager that the oldest analog profession in Washington is finally becoming a product.
The expertise question is the interesting one. Most AI startups pointed at a regulated industry are staffed by engineers who have never sat in a committee room. Fed10's founders have, repeatedly, before they were legally adults. Whether that translates into a durable moat or just a great origin story is the thing to watch. But it does mean the people building the amendment-drafting agent know what a real amendment has to survive.
The pricing story is really a distribution story. If Fed10 can deliver a consultant-grade output at a software price, it does not just win the customers who already buy lobbying - it creates new ones. The mid-sized fintech that never hired a policy team because a retainer was absurd for its size is suddenly a buyer. That is the shape of a lot of successful software: it does not steal the incumbent's customers so much as serve everyone the incumbent could never afford to reach. Whether Fed10 lands there depends on execution it has not had time to prove, but the target is coherent, and the founders can describe the customer's pain from memory rather than a market report.
The last analog profession
Most of the current wave of AI agents is chasing familiar ground - writing code, answering support tickets, booking meetings. Fed10 went somewhere stranger: the state capitol. If the thesis holds, it points at a broader pattern worth noticing. The professions ripest for automation may not be the obvious technical ones, but the expensive, opaque, relationship-heavy ones where the incumbents charge $500 an hour and still miss the committee change.
There are real reasons it might not work. Regulation moves through relationships, and relationships are stubbornly human; a staffer takes a call from someone they trust, not from an inbox. An amendment drafted by software still has to be carried by a person who understands the room. Fed10's answer is that it is not trying to remove the human from the last mile - it is trying to make sure the human shows up on the right day, with the right document, pointed at the right person. The agents do the reading and the drafting that no team can do at volume; the client does the calling. Whether that division of labor holds up under a real legislative fight is the thing the next year will decide.
There is a version of the near future where a company's entire policy team is one person and a swarm of agents watching every legislature on earth. Fed10 is building toward that version. It is early, small, and unproven at scale. But the direction is clear, and the founders have already done the human version of the job - which is more than most people automating a profession can say.