In Today’s Issue:

⚖️ Why plaintiff firms, not BigLaw, are the canary in the coal mine for AI in professional services

⏱️ What a demand letter that used to take a full day now costs in time

🤖 How Atlas, Jenny, Auditor and Analyst divide up a case file between them

🔍 What "attorney approval" really means once the machine wrote the first draft

📉 Whether the billable hour dies or simply gets repriced

🎓 His pushback on the junior-lawyer training question

🔮 The three-year prediction he is willing to put his name on

Dear Readers,

Most of the professional world still sells time. Lawyers, consultants, accountants and agencies all price their work in hours, which means the hour has to keep standing in for the value delivered. AI is quietly breaking that arrangement. When a task that used to fill a day takes minutes, an honest invoice gets smaller, and the firms with the most to lose are the ones with the least reason to move first.

Which is why the most interesting laboratory in professional services right now is not BigLaw. It is the plaintiff bar: the lawyers who sue on behalf of injured people and are paid a share of what they win rather than a rate per hour. They have no billable hour to protect, so they have been free to automate as hard as the technology allows. Whatever happens to knowledge work once the incentive to slow AI down is removed, it is happening there first.

Jay Madheswaran builds the software many of those firms are automating with. He co-founded Eve in 2023, raised a $103 million Series B at a valuation above $1 billion in September 2025, and in June 2026 launched EveOS, pitched as an operating system for a law firm rather than another tool inside one. Eve now reports more than 1,400 plaintiff firms and over 200,000 active matters on the platform.

We asked him what is genuinely automated today, what is only being sped up, and whether the billable hour survives any of it. He was more candid than the pitch deck. The billable hour is "not really my lane," hallucinations are measured largely by customers reporting them, and the bottleneck does not disappear, it moves to the humans doing the reviewing. He also pushed back, firmly, on our question about junior lawyers.

All the best,

Kim Isenberg

Eve sells software into one specific corner of the legal market: plaintiff firms, the lawyers who sue on behalf of injured people and are paid only if they win. Founded in 2023, the company has raised $164 million in total, most recently a $103 million Series B led by Spark Capital at a valuation above $1 billion, and in June 2026 it launched EveOS, a bundle of case-data, agent, audit and analytics layers it calls an AI-native operating system for a firm. That market is also the cleanest natural experiment available for a question the rest of professional services keeps postponing: what happens to a business built on billing time once the work stops taking time? We put it to Madheswaran directly. What follows is that conversation, lightly edited.

That's what Warmly proved. They defined ICP, scored buying intent, and surfaced the right accounts before a human ever touched a lead. HubSpot noticed.

On August 12, Max and Keegan are rebuilding it live in HubSpot — and showing you how to replicate it this week. HubSpot Credits included when you join HubSpot for Startups.

Grab Your Spot Now

Why are plaintiff law firms a useful early indicator for how AI may reshape professional services more broadly?

Plaintiff firms are the ‘canary in the coal mine’ for professional services because they work almost entirely on a contingency basis, meaning their incentive for adopting new technologies that help them do their job better is much clearer than for other types of firms.

What we’re seeing in this market is interesting. Two years ago the industry was essentially AI curious, but very few of them were using AI in their practice in a meaningful way. They were asking “should I be using AI?”. Today, it’s a different conversation entirely. Almost every firm is using AI in some way, and the question now is “How can I use AI to the greatest advantage?”.

We’re already seeing the most AI-native law firms substantially outcompete other firms. They’re growing faster and producing better outcomes for clients. And I think in two more years, that will be the case with other professional services firms as well.

When you talk about the end of the billable hour, what concrete change are you already seeing rather than predicting?

I can't speak much to the billable hour, since plaintiff firms run on outcomes, but I can tell you where lawyers and paralegals are spending their time is shifting dramatically.

A few years ago, even with a solid template in place, drafting a demand letter for a new case took a lawyer a full day, sometimes more. Today that's down to minutes, and the quality is higher: better command of the facts, and a figure backed by well-reasoned legal argument.

One example: a lawyer at McArthur Law Firm told us a report that used to take two days to manually review and draft was completed by Eve in seconds. Another customer is planning to scale from 25 to 200+ hearings a month using the platform.

So the marginal cost of producing a high-quality document has dropped by more than 90% in time. That frees lawyers to spend their hours elsewhere: taking more cases to trial, which we're already seeing and which could meaningfully change the legal system, or simply taking on more clients.

Productivity is shifting from essential-but-repetitive work (document creation) to essential, revenue-generating work that still requires a human.

How does contingency-based compensation change a firm’s incentive to adopt AI compared with hourly billing?

It just aligns all of the incentives perfectly. With no hourly rate coming in, a plaintiff’s side lawyer lives or dies by 3 things: 1) how many clients they take on; 2) how fast they can get the case across the finish line; and 3) how much they secure in settlement or verdict.

AI can help lawyers do all 3 of those things, so now you’re starting to see plaintiff law firms with growth rates that put most SaaS companies to shame. Frontier Law Center grew revenue by 250% in 2 years, but only added 40% more headcount.

Laurel Employment Law went from a team of 2 in 2024 to a team of 80+ with over 1300 clients in under two years. In both cases, the firms re-built their operations around AI, and reaped the rewards.

Firms who operate on a billable hour model just don’t have the same incentive to do that. It’s hard to look at a lawyer whose performance is based on their billable hours and demand they re-learn everything they’re doing to put AI at the center of their workflow — the incentive just isn’t there.

Which parts of plaintiff-side work are being fully automated today, and which are only being accelerated?

Let’s start with what can’t be automated. You’re always going to need a human in the courtroom. At least in our lifetime. Same thing for negotiation. If you’re looking to settle a case, someone has to be there in the room negotiating with the other side. So for that, AI is actually having a huge impact. We’re seeing firms using Eve repeatedly set records for the biggest verdicts in their state history, like Ricci Law’s recent $101 million verdict in North Carolina, but that is essentially a human-led task that AI can augment.

Outside of that, you can actually automate a shocking amount of the legal workflow provided you take the right steps to set things up properly.

Where must an attorney remain in the loop, even if the underlying workflow is technically automatable?

Because at the end of the day, a robot can’t appear in court. A robot isn’t licensed by the bar. Ultimately, having AI draft work for you is functionally the same as having a paralegal or another non-attorney staff member draft it. They can draft it, but the attorney has to put their name on it. So our stance has always been that Eve is here to serve the attorney, and the attorney needs to review every output for accuracy before sending it out the door. That’s not going to change any time soon, and it honestly shouldn’t.

How do Eve’s Agents, Auditor, Analyst, and underlying data layer work together across the lifecycle of a case?

Everything starts with Atlas, which is our case data layer — it's always updating. Every record, bill, call transcript, filing that comes into the firm gets extracted and structured automatically, so nobody's manually entering data for the system to know exactly where a case stands. That's the foundation everything else acts on, instead of just reporting on.

Agents sit on top of that. Jenny, our AI voice specialist, handles intake — she captures, qualifies, and signs leads 24/7. From there, agents handle records follow-ups, treatment check-ins, client communications, research, drafting. They're watching case state and acting when the conditions are right, with attorney approval built in where it matters.

Auditor is the review layer. Every night it goes through every active matter the way a senior attorney would — surfacing an injury buried in ER notes, flagging a treatment gap before demand, catching documentation that's missing. So every morning, a case manager opens up to a prioritized queue of the highest-value things to do that day, not a blank inbox.

And Analyst is the view for leadership — revenue, settlement pacing, attorney performance, caseload, all queryable in plain English, all pulled from that same live data.

What event or evidence allows an agent to start work autonomously, and what prevents it from acting on an incomplete or misunderstood case state?

Because every law firm works differently, each law firm can configure how they want their agents to work. So, you can configure an agent to start working once new records are uploaded to the matter in Eve to update a medical chronology, for example. We have a dedicated team of onboarding specialists, customer success team, and AI outcomes managers who embed with each client to set up their agents and automations to make sure that the agents are only producing work that’s actually useful. Ultimately, once firms are set up this means they spend far less time producing work and far more time taking on more clients or working on case strategy.

How do you measure hallucinations and other material errors for individual workflows such as demand letters, discovery responses, or medical-record analysis?

Inside the application our users can alert us when Eve makes a mistake or has a hallucination. Eve experiences hallucinations in a tiny, tiny fraction of the work product produced.

What does human approval mean in practice: a substantive legal review or sometimes little more than clicking “approve”?

We encourage every attorney using Eve to thoroughly read, review, and approve legal work that goes out the door with their name on it. One of the things that makes Eve so popular with our customers is the product makes it extremely easy to do that. It’s not like ChatGPT or Claude where you have to dig through verification; we’ve created a workflow for attorneys that makes this very straightforward.

You cite higher attorney capacity and settlement values. What is the denominator, comparison period, sample size, and methodology behind those figures?

Eve today works with over 1400 plaintiff firms across the United States. Any settlement value or capacity claim we make in our marketing is attributed to a specific law firm who self-reported those results to Eve, reviewed our case study about their case, and approved it.

How do you separate Eve’s contribution from selection effects, such as high-growth firms being more likely to adopt AI in the first place?

Not every firm using Eve is high-growth — we have plenty of larger, established firms using Eve who aren’t as interested in growing as they are in getting more efficient with the volume they already have. We also have lots of smaller firms using Eve who maybe don’t want to grow, but want to specialize or spend more time going to trial, for example. Ultimately you’re going to see a little bit of selection effects, but based on the reported results from our customers, Eve is having a tremendous impact on the way these firms work.

If AI lets one attorney oversee three times as many cases, how do you prevent review capacity from becoming the new bottleneck?

Funny you should ask… we’re seeing this with engineers at Eve, where our coding agents have gotten so good that code production is no longer the bottleneck; human review is. This is the likeliest outcome for firms who fully adopt Eve, and while it is still a bottleneck it’s a great problem to have. Most firms today have a bottleneck in document creation, which by its nature takes a long time. Review happens much faster — so if that’s the bottleneck, a firm is far more profitable and can hire against that need.

Junior lawyers traditionally build judgment through research, drafting, and document review. What replaces that training layer when AI performs more of the execution?

I'd push back on the premise. The repetition was never the teacher, exposure was, and AI gives juniors way more of it. An attorney at a firm running on Eve sees more cases and more variety, and gets to the work that actually builds judgment (strategy, review, figuring out what a case is worth) years earlier than they would chasing records. The core skill shifts from writing a first draft to knowing what's wrong with one, which is harder and more valuable. We see it with our own engineers: the ones who get good fastest have sharp judgment about what the AI produces, not the fastest typists. It's not automatic, firms have to put juniors in the review seat on purpose, but the ones that do build better lawyers, faster.

Does AI actually make the billable hour unsustainable, or will firms simply adapt rates, staffing, and fee structures while keeping it?

As I mentioned earlier, the billable hour isn't really my lane, since the firms we work with are on contingency. But my read is that AI doesn't kill it outright, it just strains the logic. When work that used to fill a day takes minutes, the hour stops tracking the value being delivered, and clients start asking why. So firms adapt before they abandon it: more flat fees, more value-based pricing, repricing what's left. Whether it survives that is a question for the defense bar. What I can tell you is the firms we work with never had that tension to resolve, and it's a big reason they can put AI at the center of how they operate without fighting their own comp model.

What conditions must exist for the plaintiff-law model to transfer to consulting, accounting, or research, and where does the analogy break down?

Two things. The firm can't be paid by the hour, because that's what kills the incentive to automate. Plaintiff firms live on outcomes, so every hour AI gives back is pure upside. Consulting and accounting still bill for time, and that's a big reason they've been slower to move. And the work has to be repeatable enough that a system can actually do it, with a human owning the call that matters. Where the analogy breaks is accountability. In our world, a licensed attorney signs the work and carries the liability, so there's a review point you can't skip. A lot of consulting doesn't have that clean line, and where it's fuzzy, adoption gets messier. The fields that look most like plaintiff law (paid on outcomes, with a licensed human on the hook) are the ones I'd bet move next. Long term, I do think that most professional services will move to an outcomes-based model. You won’t pay your accountant by the hour, you’ll pay them a fixed rate to complete the monthly close and file the taxes.

What measurable prediction about AI and professional-services economics are you willing to make for the next three years?

Here's one I'll put my name on. Within three years, the gap between AI-native plaintiff firms and everyone else will be too wide to close by hiring, because you can't out-recruit a competitor who's decoupled growth from headcount. We're already seeing it: Frontier grew revenue 250% on 40% more headcount, Laurel went from two people to over eighty. Today that's a handful of standout firms. In three years it's the baseline, and the firms that didn't rebuild around AI are the ones explaining to clients why they're slower and more expensive. And I'd bet we're having this same conversation about accounting and consulting soon after.

This interview was conducted by email and has been edited for length and clarity.

The best voice models, now with full orchestration. Build real-time voice and chat agents on one low-latency stack: any LLM, your tools and knowledge, testing, Guardrails, and omnichannel deployment.

Start building free

The most useful thing Madheswaran says is also the thing his own product cannot fix. The billable hour is not being broken by capability, it is being broken by incentive, and he is candid that his customers never had the problem to begin with: plaintiff firms are paid out of the recovery, so every hour AI hands back is margin rather than lost revenue. That is why the fastest adoption in law is happening at the end of the market with the least prestige and the least to protect. For everyone still selling time, the uncomfortable part of his answer is that the capability arrives long before the willingness to reprice does. "When work that used to fill a day takes minutes, the hour stops tracking the value being delivered, and clients start asking why."

The second idea worth carrying out of this conversation is where the constraint goes. Madheswaran does not claim it disappears. He says it moves to review, and he offers his own engineering team as the evidence: at Eve, code production stopped being the bottleneck and human review became it. A firm that triples its drafting capacity has not removed a limit, it has relocated it onto the one person who has to put their name on the work. That is a better problem than the one it replaces, and it is still a real one.

Two caveats belong on the record. Every customer figure in this interview is self-reported by the firm and approved by Eve, which he says plainly when we asked about methodology, and Atlas and Analyst were both still in customer beta at the June 2026 launch, so parts of the system described here are further along in the pitch than in the product. Our thanks to Jay Madheswaran and the Eve team for the exclusive.

Sources:

🔗 Eve, "Eve Launches EveOS, the AI-Native Operating System Transforming Plaintiff Law Firms" (11 June 2026): https://www.businesswire.com/news/home/20260611528924/en/Eve-Launches-EveOS-the-AI-Native-Operating-System-Transforming-Plaintiff-Law-Firms

🔗 Eve, "Eve Raises $103 Million at $1 Billion Valuation" (30 September 2025): https://www.prnewswire.com/news-releases/eve-raises-103-million-at-1-billion-valuation-to-help-plaintiff-firms-deliver-justice-through-ai-transformation-302570807.html

🔗 Eve, Frontier Law Center case study: https://www.eve.legal/case-studies/frontier-law-center-case-study

🔗 Eve, Laurel Employment Law case study: https://www.eve.legal/case-studies/laurel-employment-law

🔗 "Landmark $101 Million Verdict Awarded in North Carolina Personal Injury Lawsuit" (National Law Review): https://natlawreview.com/press-releases/landmark-101-million-verdict-awarded-north-carolina-personal-injury-lawsuit