The Compression of Consulting: What Survives When AI Takes the Scaffolding

The conversation about AI and consulting tends to split into two camps. One argues that senior advisory work is about to be commoditized. The other argues that nothing important will change, because AI cannot replicate experience. Both are wrong in a way that matters for anyone buying or delivering advisory services over the next few years.

What is actually happening is more specific than either camp suggests. AI is not taking the judgement. It is taking the scaffolding that judgement used to stand on.

I should be direct about where I am writing from. I have been advising and operating at board, executive, and PE level for over two decades, and the last eighteen months have been the first period in which I have seen this compression move at speed inside live engagements. What follows is a description of what it looks like from inside that work — not a forecast. The tone is diagnostic rather than dramatic because the situation, closely observed, has a predictable shape and a workable path through. Most of this essay is spent describing the shape. The path through is the subject of the essays that follow.

What the scaffolding actually is

A traditional strategy engagement does not consist primarily of senior judgement. It consists of the analytical machinery that surrounds senior judgement and makes it presentable: benchmarking, peer comparisons, process mapping, stakeholder interviews, synthesis documents, committee-ready slides, status reports, and the week-by-week cadence of meetings that converts raw material into conclusions.

This analytical support for organizing and delivering senior judgment is labor-intensive, and it is what makes a twelve-week diagnostic cost what it costs. It is also, very specifically, the thing that large language models and their adjacent tools are now genuinely good at. A small team equipped with current tools can produce in four weeks what used to take a team of four over twelve. That is not a prediction. It is already true in several of my engagements this year.

This has two notable downstream effects. It compresses the time-to-insight of any well-executed engagement. And it shrinks the pricing of any engagement component for which the primary value was the scaffolding itself.

The compression is uneven, and that is the point

The market for consulting is not one market. It is at least three, each with very different exposure to what is happening.

The first is the market for process work — PMO support, program delivery, standardized assessments, benchmarking studies, and the industrial production of decks. This market will compress hard and quickly. The combination of AI-augmented junior consultants and purpose-built tools will continue to lower the labour cost of producing this kind of output, and clients will notice.

The second is the market for generic transformation advisory — the broad category of senior directors who run cross-functional programs, facilitate leadership workshops, and translate strategy into execution plans. This market will not disappear, but its economics will compress meaningfully. The generalist currently in healthy demand will find that a well-equipped operator can produce comparable output in less time, and that prospects are increasingly able to tell the difference.

The third is the market for structural diagnosis and irreversible judgement. Here the economics move in the opposite direction. As AI proliferates inside client organizations, the premium attached to a senior human whose judgement is on the line does not diminish. It increases. The reason is mechanical: the more systems in an organization operate at machine speed, and the more of them make decisions whose logic no one can fully reconstruct, the more the organization needs a human who can look at the situation and say what is actually happening and what should be done about it.

The compression, in other words, runs through the middle of the market and leaves both ends intact for different reasons. The low end is protected by price. The high end is protected by accountability. For clients operating at the senior end of this market, the practical implication is encouraging: the advisors who survive this compression are, by construction, the ones whose value was never in the scaffolding in the first place.

Four capabilities AI makes more valuable

It is worth naming these precisely, because they are the foundation of whatever advisory practice survives the next five years — and, equally, of what a well-run organization should be buying from its advisors in a world where scaffolding is no longer scarce.

Accountability for irreversible decisions. An AI system can model every scenario around cancelling a €40M program. It cannot — should not — sign off on cancelling it. It cannot sit in the board meeting and say, with a reputation at stake, that this is the right thing to do. Someone has to take accountability for calls that cannot be undone. The more sophisticated the analytical environment becomes, the more visible this separation becomes — between what a system can calculate and what a human has to decide. Organizations that pay for this capability do not get less of it in the AI era. They get more access to senior attention, because the scaffolding around it has become cheaper to produce.

Political navigation in complex organizations. The engagements I remember most clearly did not succeed because of analytical superiority. They succeeded because someone was able to read the power dynamics in a room, manage the subtle relationships between people whose incentives were not aligned, and move a complex human structure from one state to another. The stabilization of a failing program at a Global Financial Services Group, a scope reset at a Regulated Infrastructure Operator, the founder-adjacent work at a Biotech Scale-Up — each depended on a kind of situational reading that no model performs. The ability to be in the room, to know what should be said and what is best not, and to take responsibility for expressing it, is not commoditizing.

Economic translation of technical risk. One of the more durable pieces of work I have done was converting roughly 500 IT and operational risks into a single economic exposure view for a Global Insurance Group. It was useful not because the underlying data was novel — most of it was already known inside the organization — but because it rendered technical complexity into language a board could act on. As organizations deploy AI more broadly, the demand for this kind of translation grows. Boards are about to face AI-specific risks they cannot interpret directly. The work of executives who can translate system behavior into governance and financial language is not becoming trivialized. That market is expanding.

The trust-based confidential diagnostic. There is a particular kind of relationship, normally undocumented, between an executive and a senior independent figure who can be told what is actually going on. This relationship has no AI equivalent. The more autonomous systems an organization deploys, the more executives need a human who sits entirely outside the machine and whose judgement is not mediated by any of it. This is the oldest form of advisory work, and it is probably the one most insulated from the current wave of change.

The shift in what is being bought

The quiet implication, for anyone operating in this market, is that the unit of value is changing.

Through most of the last three decades, clients bought process. They bought engagement timelines, deliverables, methodologies, and the reassurance of visible effort. The logistics was not incidental to what they were buying — it was a significant part of it. A board or an executive committee could look at the volume and quality of the work and feel that they had bought something real.

That link is breaking. Scaffolding is cheaper to produce and therefore harder to charge for. What remains chargeable — and what, in fact, is becoming more valuable — is a narrower and more demanding thing: the structural diagnosis of a situation, and the willingness to make calls that cannot be undone. Clients who understand this will shift their buying behavior. They will pay less for production and more for decision-making insight. They will care less about how many interviews were conducted and more about whether someone can tell them, in clear language, what the real constraint is and what to do about it.

For those already operating in this market, this is not bad news. It is the market finally aligning with the core value that was always being delivered. For clients, it is an opportunity to pay for the thing that was actually useful all along, and less for the apparatus around it.

A note on navigating this

None of what this essay describes is new in kind. Markets have absorbed compressive technology shifts before, and the organizations and operators who navigated them well had three things in common: they named the shift early, they understood which part of their value was genuinely durable, and they adjusted the shape of what they sold before the market forced the adjustment on them. The compression now underway in advisory work has a predictable shape. Organizations that want to buy wisely through it, and advisors who want to stay relevant through it, benefit from the same disciplines — clarity about what scaffolding is worth, clarity about what judgement is worth, and a willingness to redesign the engagement around the latter.

That is the work. It is difficult, but it is not mysterious.

What this series is about

This is the first in a short series of essays in which I explore what AI is doing to the work my clients aim to govern, and to my own work as I seek to serve them.

The next essay turns to a specific failure mode I am now seeing consistently inside large organizations that have moved past their initial AI experiments: the uncoordinated proliferation of autonomous systems across the enterprise, and what it takes to govern it well. This is a problem with a clear shape and a workable response — which is, ultimately, what this series is about.