Pattern Counsel is written for General Counsel, Chief Legal Officers and their teams—and for the law firms, legal operations professionals and technology providers who work alongside them. Originally published on LinkedIn. This work is AI-assisted—read about how I work.

Introduction

Legal is an enterprise resource, not a service provider.

It compounds value through embeddedness — in the business and across the phases of production that demand it. It trades replicable work out to a governed process to maximise compounding value. And it is judged by what the enterprise achieves where legal is embedded, not by what legal produces.

That is not a theory I built. It is a pattern I recognised, late, after fifteen years running a legal function while it was being cut. I am not certain it is new, and I don’t need it to be. I arrived at it from the inside, through a long contraction, and I am setting it down now for one reason.

The same pressure my teams faced is reaching the entire profession again, but now from a different direction. The CFO asking the legal team to do more with AI is not, at the level that matters, asking for better law. The word is “improvement.” The instruction beneath it is to spend less on this and relieve the business’s dependence on it.

My version of that pressure came from a commodity cycle, not a technology. The cause was nothing alike. But the shape of the pressure was the same. And for in-house legal teams across industries, there is no cycle to wait out this time. With the arrival of AI, the cost of sourcing legal support is not dipping; it is resetting.

So I’m setting it down here, in parts, as a newsletter. Each instalment takes one piece of the pattern and reads it out. This is not a model to adopt; it’s the shape of what one function found it could not give away. And it’s now being offered to a profession that’s being asked the very same questions.

Here is where those patterns came from.

The story

I built the South African and later global upstream legal functions of a global diversified mining company from 2010, integrating lawyers drawn from diverse teams that had run separately, with their own habits and loyalties. Each time I took on this challenge, the early work was exhilarating and focused on construction.

Then inevitably, the moment would come when the economics or the group strategy would shift, and the work would become subtraction.

What followed, for the better part of one and a half decades, was retrenchment, in both senses of the word. The business would shed people, and the function that advised on the shedding would be under the same pressure to shrink. Every cycle, the question returned, in one form or another. Why does this cost what it costs, and what happens if there is less of it?

That’s my story, but I have no doubt it’s been your story too.

I never found clean answers to that pressure when it was at its worst. But I eventually found four patterns that emerged within those cycles. I offer them as field notes rather than theories or models.

First pattern: a resource is not a service.

The pressure to cut is indiscriminate when it arrives at scale.

I have seen mining companies cut engineers, investment banks cut risk analysts, life insurers cut actuaries. Nothing is sacred. But there is a distinction worth holding, even if it offers no immunity.

When a mining company cuts engineers because it is producing less, it is making a production decision — the business needs fewer because it is doing less. When a business cuts its legal function because it believes the work can be sourced more cheaply outside — through a law firm, an outsourcer, an AI-assisted platform — it is making a sourcing decision dressed as a capacity decision. The legal work has not contracted. The business has concluded it is paying for a service it can buy differently. Legal is especially easy to miscategorise this way, because the work is intangible and the sourcing alternatives are multiplying.

The function that accepts the service-provider description has confirmed the sourcing logic. Every efficiency it delivers becomes evidence that more can be extracted. Every technology it adopts becomes the justification for the next reduction. The contraction is not a risk it might lose. It is the trajectory the framing invites.

What the business loses when it reduces legal on sourcing grounds is not volume. It is position: the embedded judgment that cannot be bought back in when the exposure arrives. You cannot rehire the accumulated knowledge of how this particular enterprise makes its decisions, which risks it has actually lived, what its regulators have said and not said over twenty years. That is not on any market.

A business does not replace its own judgment with a cheaper version of itself. It replaces services. Legal’s claim — the only one that holds under sustained pressure — is that it is the former, not the latter.

Second pattern: value compounds, and only from inside.

The obvious answer to what makes an in-house legal function valuable is the wrong one. It is not legal expertise. Legal expertise is everywhere, and a well-trained large language model now holds more legal information, instantly, than any team ever will. If information were the asset, the work would have left already.

What being inside the business produces is a different kind of knowledge, and it has no market. How decisions actually get made here, as against how the policy says they do. Which risks turn real and which only sit on a register, year after year. Who in the room can move a decision and who merely attends. None of that is in any document. It accumulates in the people who stay long enough to watch the same patterns return.

And it compounds. Each cycle of presence sharpens the next judgment, because the next judgment is made by someone who was there when the last one landed. An adviser reading the file afterwards reads what was written. The person who was inside reads that, and everything around it that never reached the page.

There was a lawyer on a team I once led who was accountable for mining, environmental law, and permitting. A tailings dam is built in stages, raised in successive lifts, with the permitting regime requiring each new lift to be separately licensed before the next can begin. She knew every permit through every cycle of its evolution — why each prior lift had been approved the way it was, where the regulator had pressed, what the record did and did not say. That is the compounding asset made literal: a structure that physically accumulates, where the licence for this lift rests on the history of the last, read by someone who had lived the whole climb.

But the memory was only the first part of it. She sat on the permitting governance committee — often co-chaired it — accountable alongside the engineers, the hydrologists, and the community team for whether the permit landed or failed. She was part of a legal team led by a senior lawyer who sat on the business Exco. Three layers, each doing its part. None of it reducible to a person who happened to be good.

Her presence wasn’t uniform across the cycle. At the front, when each lift’s application was being designed, she was one of the few whose contributions were most consequential. In the execution phase, all disciplines contributed alongside each other — she was one of many. At the back, when the outcome was reviewed and the next application’s strategy shaped, she was one of the few again. Front and back: that is where the compounding happened.

She knew which questions to ask at each lift because she had been there for the ones before. Did the hydrologist’s analysis in this cycle account for what the record showed in the last. How did that administrative decision — the one that read as procedural at the time — carry into the licence conditions being built on now. An external adviser instructed at this lift does not know which of those questions matter, because they were not there when the answers formed. She was.

That diagnostic precision, knowing the shape of what might have been missed, across every discipline, across every cycle, is not in any record. It lives in the person who has been present and is accountable for what comes next. There are no mines without permits.

The rejoinder is that artificial intelligence changes this. A model can read every contract, every board paper, every permit ever filed. But it accumulates the record. The record is what was decided. Judgment is the call about what to do next, made by someone who can be held to account for it, who reads the room as well as the file, and who carries the cost when it goes wrong. A system can hold the whole history of the enterprise and still not own a single decision.

Accumulation is not judgment, and it does not become judgment by getting larger.

That judgment, to mean anything, has to be present at the moments that shape what the enterprise does next. Every stream of work the enterprise undertakes has the same three moments: a planning phase, where what will be done and how is determined; an execution phase, where it is carried out; and a review phase, where outcomes are assessed and lessons carried into the next plan. A market entry, a major contract, a regulatory response, a restructuring — the enterprise moves through that shape every time. Legal’s compounding happens when it is present inside the planning and the review, not summoned only after the plan is already set and execution is underway. The question worth asking is at which moment your function typically arrives — and whether that matches where the value actually accumulates.

Third pattern: trade the replicable out, but keep the governing

Not all work compounds. The work that doesn’t is the work whose value doesn’t depend on being inside — work where an external provider, properly instructed, would reach the same result. It leaves nothing behind that sharpens the next pass. The enterprise loses nothing by sourcing it from outside, because being inside added nothing to it. That is the work to trade out.

Give it away. The capacity that frees is not spent going faster; it is spent where value compounds.

Delivery is divestible. The oversight is not.

I learned the limit of that the hard way. The moment we sent the managing of the work out along with the doing of it, we had not saved capacity. We had lost hold of it. Drawing the boundary of what the external channel does, routing each matter to the right channel, catching the error that replicates silently across a thousand instances before anyone reads them: these are not coordination tasks. They are embedded judgments, and they belong inside.

There is a decision underneath this that I will return to in a later instalment — which streams of work warrant keeping inside, and which are better sourced. The axis that matters is not the difficulty of the work. It is whether being inside changes the outcome, and whether presence across many instances builds something that compounds. That question, and the decision it drives, deserves its own treatment.

Fourth pattern: judged by the enterprise, not the department.

Legal’s contribution is measured by what the enterprise achieves where legal is embedded — not by the volume or efficiency of the legal work it produces.

This is not an argument against measurement. It is an argument against mismeasurement.

The function evidences its contribution. It does not reduce it to a figure.

The value resists pricing, and honestly so. It is a stock that builds invisibly and shows mostly in absence — the loss that did not occur, the exposure caught before it formed, the decision shaped at the front of the cycle rather than contested at the back. Activity metrics fail here not because they are inaccurate but because they count motion while the asset is stock. The harder I pushed legal risk to translate into what the business already knew how to count, the more it came to look like counting motion. I watched good lawyers before me lose that argument by accepting the wrong scoreboard, and I understand why they accepted it. The pressure for a number is relentless. But a function that answers with a number is confirming that a number is what it is worth. It isn’t. It is worth what the enterprise achieves, in the work it was embedded in.

That is not a soft claim. It is a harder one. It asks the function to stand behind outcomes rather than outputs, and to resist the safety of metrics that prove only that it was busy.

Concluding this instalment.

Four patterns. None designed. All arrived at by watching what could not be removed without something being quietly lost that could not be bought back.

The instalments that follow take each one further — the full logic of the resource distinction, the epistemology of embedded knowledge and why AI sharpens rather than dissolves it, the discipline of determining where presence is demanded and where it isn’t, and the honest architecture of measurement.

What was still standing when it had done its work is what the rest of this sets out.

A note on sources

This essay emerged from experience rather than research, but experience and reading rarely stay separate for long. Over fifteen years inside a function that was repeatedly asked to justify its existence, I read widely about where legal thought leadership was taking the profession. Some of that reading sharpened ideas I was already forming; some of it named things I had felt but not yet articulated. I cannot always say which is which.

Simmons and Dinnage wrote seriously about why continuous presence inside the enterprise creates value that episodic engagement cannot replicate — and gave me a way to name what I had been watching. Ben Heineman’s case for the lawyer-statesman, set out in The Inside Counsel Revolution, shaped how I thought about what a general counsel is actually for. Deming and Shewhart’s improvement cycle gave me a convenient way to name the phases of work, even if I have borrowed the lens without adopting the whole frame. And CLOC’s emerging standard for legal operations measurement is the reference point against which the measurement argument in this essay pushes back: not because the metrics are wrong, but because they count motion while the asset is stock.

The perception gap between general counsel and the executives they serve has proven remarkably persistent — and has, if anything, widened. Bain first documented it in 2010, finding that almost forty percent of CLOs rated their function highly for commercial value while only fourteen percent of CEOs agreed. Sixteen years later, Thomson Reuters found the gap had grown: eighty-six percent of general counsel believed their department contributed significantly to organisational objectives, while only seventeen percent of the C-suite agreed — and forty-two percent said legal contributed little or nothing at all. That is not a communication failure. It is a framing one. And it is what put a number on something I had lived without being able to prove.

I offer these acknowledgements not as a bibliography but as an honest account of how thinking forms — from pressure, from practice, and from the writers who happened to be working in the same space.

Continue the Discussion

If these patterns resonate with your own experience, or if your experience points in a different direction, I would genuinely value the discussion.

Kevin Lester is an independent counsel and thinking partner. He spent fifteen years as a Country Head of Legal and subsequently as General Counsel at Anglo American. He advises General Counsel, senior executives, and boards at kevinlester.co.za.