The "neofirm" has a people problem

What happens to the practitioners when the craft gets automated

Ryan Daniels published a piece recently arguing that professional services firms are having their Bessemer steel moment. The structural argument is sharp: partnership models pay out every dollar as compensation, leaving nothing for R&D. They bill by the hour, which penalizes efficiency. They can’t sell equity to outside investors, which prevents long-term investment. The “neofirm” model he describes, equal parts practitioners and engineers, outcome-based pricing, obsessive quality measurement, is a genuine advance over the century-old partnership structure.

I buy the business model argument. The incentive analysis is clean: when you bill by the hour, you’re selling labor. When you bill by the outcome, you’re selling value. SpaceX moved from cost-plus to fixed-price and redefined what was possible. Neofirms will do something analogous in law, consulting, and accounting.

But there’s a sentence buried in the piece that reveals the problem nobody in this conversation is solving. Daniels writes: “Every hour a lawyer spends experimenting with an engineer is time they could have spent working on a new client’s request. It’s lost revenue, and we argue about how to allocate lawyer time constantly.”

That argument about time allocation is the most important thing in the article. Not because it’s a resource management problem. Because it’s an identity crisis wearing a calendar.

The practitioner’s proof system just stopped compiling

A partner at a top law firm has spent twenty years building expertise. They can read a contract and spot the risk that a junior associate would miss. They can navigate a negotiation because they’ve been in the room for hundreds of them. They can advise a client on a bet-the-company decision because they’ve seen what happens when the bet goes wrong.

Their identity is built on this expertise. It’s how they evaluate themselves, how the market evaluates them, and how they justify their $3,000-an-hour rate. The proof loop is tight: I know things others don’t, my knowledge produces better outcomes, the outcomes justify the fee.

Now the neofirm arrives and says: we’re going to automate everything beneath the automation threshold, and the threshold keeps moving upward. Contract review? Automated. Due diligence? Automated. First-draft negotiation? Agent-to-agent. The partner’s job is now to define the benchmarks, guide the engineers, and handle the edge cases that AI can’t.

This is technically correct. It’s also a complete redefinition of what “good” looks like in a career that person spent two decades mastering.

I’ve written about this mechanism repeatedly: when the environment changes faster than a professional’s self-concept can adapt, the result is identity disruption. The metrics you used to evaluate yourself just got invalidated. Not because you got worse. Because the definition of value shifted. The partner who could read a contract better than anyone in the room is still the best reader in the room. But the room now has an AI that reads 500 contracts an hour and flags the same risks with 94% accuracy. The partner’s advantage isn’t gone. It’s narrowed to the 6% the AI misses. That’s genuinely valuable. It doesn’t feel like it.

“Your role is elevated” is not a job description

The standard comfort for practitioners in this transition is: AI handles the routine work, you handle the judgment. Your role isn’t diminished, it’s elevated. You’re freed from drudgery to focus on the highest-value problems.

I wrote about why this framing is psychologically insufficient in the context of engineering. It applies with equal force to professional services.

The “routine work” wasn’t just revenue generation. It was the activity through which the practitioner maintained their expertise, their pattern recognition, and their sense of professional identity. A lawyer who reviews contracts for twenty years builds an intuition that operates beneath conscious analysis. That intuition was trained by the repetition. Remove the repetition, and you’ve removed the training mechanism for the judgment you’re now asking them to exercise exclusively.

This is the same finding Steve Yegge described in his AI Vampire piece: AI concentrates the hard cognitive work (judgment, decisions, quality evaluation) and eliminates the easier work (production, drafting, routine analysis) that provided cognitive recovery. A role that’s all judgment and no production has a different sustainability profile. Yegge’s conclusion was that the workday needs to shorten to three or four hours. The neofirm model, which measures outcomes rather than hours, could theoretically accommodate this. But I haven’t seen anyone in the neofirm conversation acknowledge that the practitioners themselves may not be able to sustain eight hours of concentrated judgment work even if the business model supports it.

The pyramid problem is an identity problem

Daniels doesn’t address this directly, but the research on professional services transformation reveals a deeper structural issue: AI is dismantling the apprenticeship model that produced the very expertise the neofirm depends on.

Traditional firms operate on a pyramid. Junior associates do the high-volume routine work. Through that work, they learn the patterns, develop judgment, and eventually become the senior partners who handle the complex cases. The routine work is both revenue and training.

AI automates the routine work. The pyramid’s base collapses. The Big Four have already cut graduate intakes by 6-30%, and graduate job postings in accounting and consulting dropped 44% year-on-year by 2024.

Here’s the question nobody is answering: if the routine work is where expertise was built, and the routine work is now automated, where does the next generation of expert practitioners come from?

The neofirm model assumes a supply of “elite partners” who can be “leveraged with armies of agents.” But those elite partners were forged by decades of doing the work that’s now being handed to AI. If you cut off the pipeline that produces them, you’re consuming a non-renewable resource. The neofirm is optimized for the current generation of experts. It has no mechanism for producing the next one.

This is the same pattern I described in the context of teams that split in half: AI adoption creates a capability gap between those who developed expertise the old way and those who are entering the profession under the new model. The old way produced deep pattern recognition through repetitive exposure to real cases. The new way produces... what, exactly? That’s an open question, and the neofirm model doesn’t answer it.

The measurement revolution is real, and it has a shadow

The most compelling part of Daniels’ argument is about quality measurement. Traditional professional services operated on reputation and prestige. Neofirms, he argues, will prove their competence through benchmarks and evaluations. “Instead of choosing your banker based on their alma mater or firm prestige, you will ask to see their benchmarks.”

This is a genuine advance. I’ve written about why explicit evaluation beats vibes-based assessment across every kind of organization. Making quality measurable is how you escape the politics of reputation and move toward something defensible.

But measurement has a shadow. When you measure output quality and automate everything beneath the threshold, you create relentless pressure to raise the threshold. The practitioner’s value is now defined as “the work above what AI can do.” Every time AI gets better, that band narrows. Every time the band narrows, the practitioner’s identity compresses.

This is the Dunning-Kruger problem in reverse. The most expert practitioners can see the gap between what AI produces and what excellence requires. They know the 6% matters. But the measurement system may not capture that 6% because the benchmarks were designed to evaluate whether the AI’s output meets a threshold, not whether the human’s contribution exceeds it. If the benchmark can’t distinguish between good-enough AI output and genuinely excellent human-augmented output, the practitioner’s edge becomes invisible to the system they’re operating in.

What the neofirm needs to build alongside the technology

Daniels is right that the business model has to change. Billing by the hour is a dead end. Partnership structures that can’t invest in R&D are a dead end. Outcome-based pricing aligned with quality measurement is the right direction.

But the technology transition is the easy part. The hard part is the human transition. And the neofirm conversation is almost entirely silent on it.

Define the new version of “good.” If the practitioner’s job is no longer producing the work product but guiding the AI and handling the exceptions, what does excellence look like in that role? Write it down. Not in aspirational terms. In operational ones. What does a great day look like for a lawyer at a neofirm? What does a great quarter look like? If you can’t answer that, your best practitioners are going to default to the old metrics and feel like they’re failing.

Invest in the identity transition, not just the tooling. The practitioners who will thrive in the neofirm model are the ones who can rebuild their professional identity around judgment, quality design, and AI collaboration rather than around the craft they spent decades mastering. That transition doesn’t happen by itself. It requires acknowledgment that the old identity is being lost, space to grieve it, and a new narrative about what expertise means now. I know that sounds soft for a business-model piece. It’s not. It’s the difference between retaining your best people and watching them walk out over the next eighteen months because the job no longer feels like the profession they signed up for.

Solve the pipeline problem before the current experts retire. If the routine work was the training ground, and the routine work is automated, you need a new training ground. Simulation? Structured AI-augmented apprenticeship? Deliberate exposure to the edge cases that AI gets wrong? Whatever the answer is, it needs to be designed now, because the next generation of elite partners isn’t going to emerge from a pipeline that no longer exists.

Signals

You’ll know the neofirm transition is healthy when the practitioners describe their work with pride rather than loss. When “I guide the AI and handle the exceptions” feels like a professional identity, not a consolation prize. When the benchmarks capture the human contribution, not just the automated output. When junior professionals can describe a credible path to expertise that doesn’t depend on doing work that no longer exists.

The neofirm is the right business model. The Bessemer analogy is apt. The incentive structure is better. The measurement revolution is real. But business models don’t experience identity crises. People do. And the people inside this transition, the lawyers, the consultants, the accountants who built their careers on a craft that’s being redefined underneath them, are the load-bearing wall that the neofirm model is built on.

If you don’t take care of the wall, the building doesn’t stand. No matter how well-designed the architecture is.