Ask a managing partner to write down the rules by which their firm operates and you get one of two answers. The honest one is “I’m not sure I could.” The other is the firm’s manual — a real document that everyone knows is not actually how the firm runs day to day. There is the manual, and there is what people actually do, and the space between them is where the firm really runs. This is not a criticism; it is how professional-services firms have always worked. It is also the thing AI cannot work with.
What the decision layer means
The decision (or policy) layer is the set of rules a system uses to decide what to do, what to escalate to a human, and what to refuse. It is the difference between an autonomous system that runs your firm into a regulatory wall and one that knows when to stop and ask. In a firm it looks like answers to questions most firms cannot answer today: What can a manager sign off without a partner — by client size, engagement type, risk? What can a senior do without manager review? When does an unpaid invoice become a stop-work decision, and who makes the call? What is the firm’s position on crypto, cannabis or politically-exposed clients — written down, or “we know it when we see it”? When a junior spots a money-laundering signal, what happens, and at what point does the firm file? These are routine questions a firm faces several times a year, and most cannot answer them in writing in a way two partners would agree on.
Why this is the hardest part of the series
The sensor layer is mostly infrastructure and partners come around to it. The decision layer is different: it demands the rules be made explicit, and the moment you start writing them down, several uncomfortable things happen at once. The partners discover they disagree with each other. Managers realise the firm has been less coherent than it presented itself as being. Some of the firm’s prized flexibility goes away, because discretion is part of how partners define seniority. And the firm discovers it has been carrying risk it never properly looked at — the clients on the books who don’t fit the rules, the historical cases that probably should have been filed. That housekeeping is a feature, not a bug, though it doesn’t feel like one in the moment.
This is not really an AI problem
AI is forcing the conversation, but the gap between the manual and reality predates AI by a century. It is unsustainable for two reasons that have nothing to do with AI. Succession: every firm losing a senior partner loses a decision layer nobody else can reproduce, and the loss is unmanaged because the layer was never explicit. And scale: the partner-knowledge model hits a wall between fifty and two hundred staff. AI just brings the deadline forward — a firm that wants to deploy it on client work has to answer the questions explicitly, because the system cannot read the partner’s mind.
Drowning in manual, repetitive work? Tell us the task and we’ll show you what to automate.
What becomes possible
A firm with an explicit decision layer onboards staff faster, because the rules are readable rather than absorbed by osmosis. Partner meetings become about substance rather than precedent. Risk is honestly distributed — the firm can see, in writing, what it has decided is acceptable. And AI becomes deployable on real work, because “do this when the engagement matches these criteria, escalate when it matches these, refuse when it matches these” is something a system can act on and a partner can sign off.
What should stay judgement
The point is not to codify everything — a firm that does produces one that cannot respond to the unusual situation, which is often the one that matters most. The trick is knowing what to codify and what to leave to judgement, and the honest answer for most firms is that far more could be codified than currently is, with a smaller residue of genuinely contextual judgement than partners tend to claim. If a new partner joined tomorrow, could they read a single document and know how the firm makes decisions? If not, the decision layer is the work.
Part 4 of The Self-Improving Firm. Daniel Lawrence is the CEO and co-founder of Bots For That.
