Consulting firms are building the wrong muscle for AI

Most takes on the future of consulting converge on the same idea: AI commoditises analysis, slideware loses value, clients expect outcomes. The conclusion seems obvious. Consulting firms need to move closer to execution.

That is directionally right. But it misses the deeper shift already underway.

The rate problem #

AI is not just changing how work is done. It is changing the rate at which change is proposed.

More code proposals. More pricing scenarios. More risk model variants. More workflow configurations entering delivery across the enterprise. But the rest of the system does not accelerate at the same rate. Validation still takes time. Integration still breaks. Governance still relies on review and interpretation.

So the constraint moves. Not to strategy. Not to engineering capacity. But to something most organisations never had to design explicitly: the ability to turn decisions into safe operational reality at increasing speed.

This is what absorption looks like at the system level. Not the capacity to do more work, but the capacity to accept change under control, with evidence, at the pace change now arrives.

Strategic decisions remain human, political, and uncertain. That does not change. What changes is the system that converts those decisions into operational reality once they exist. For years, the hard part was generating good ideas. Now the hard part is accepting them safely.

Five things that used to be implicit #

Consulting sat upstream of execution because the boundary between decision and implementation was relatively stable. AI is dissolving that boundary. Decisions now arrive faster than systems can absorb them.

When absorption fails, the symptoms are familiar: growing review queues, increasing rework, late discovery of risk, rising audit pressure. More execution does not solve this. It often amplifies it.

The structural reason is that five things which used to be handled implicitly, through experience, coordination, and organisational memory, now need to be designed into the system itself.

Most organisations still treat intent as something informal channels can carry. A well-written ticket for the engineering team. A pricing memo for the product team. A policy interpretation passed verbally to the compliance function. At human speed, that worked. At machine speed, intent that is not explicit and stable enough to govern produces drift with every acceleration. Downstream execution reinterprets what was meant, and the reinterpretation compounds.

Exploration and execution have always coexisted, but the boundary between them was never a formal design decision. It was a cultural norm. Teams knew when to stop exploring and start building. That judgement does not transfer to machine-speed delivery. Where ambiguity is allowed and where it must stop becomes a system design question, not a team maturity question.

Authority follows the same pattern. Who can approve a design change? Who authorises a new pricing structure? Who signs off on a risk model update? In most organisations, these questions are answered by role conventions and implicit trust. When AI agents participate in delivery, and when the volume of proposed changes rises across domains, implicit authority becomes a structural vulnerability. Authority that is not declared, scoped, and auditable is authority that cannot be governed.

Validation in most organisations still follows the traditional pattern: build first, review later. At human speed, the cost of late review was manageable. At machine speed, it is not. When validation runs after execution rather than during it, rejection rates climb and rework multiplies. The review queue becomes the bottleneck that was supposed to be the safety net.

Audit evidence tells the same story. In most organisations it is assembled retrospectively: someone collects artefacts after the fact to demonstrate compliance. When delivery runs at machine speed, retrospective evidence collection breaks. The change set is too large, too fast, and too distributed. Evidence needs to be emitted continuously, as a structural property of the system, not a separate administrative activity.

Each of these was survivable as implicit practise when the rate of change was low. Each becomes a failure mode when the rate of change is high.

Software delivery is the first domain where this pressure becomes visible, because its artefacts, handoffs, and failures are already relatively formalised. But the same pattern is appearing across pricing, risk, compliance, product configuration, and operational workflow automation. Wherever decisions become system behaviour, the same five implicit things break in the same way.

Where the real shift is happening #

The emerging role is not doing more execution. It is shaping how execution happens.

A pattern is becoming visible across organisations that are absorbing AI-driven change without the familiar symptoms. In these organisations, exploration is fast but bounded. Execution is automated but controlled. Governance is embedded in the flow, not overlaid after the fact. Outcomes are reproducible, not dependent on individual interpretation.

This is less about adding capability. It is about removing ambiguity from how change becomes real.

What is emerging is that absorption may be the root capability, not one of several desirable outcomes. Resilience, auditability, adaptability, compounding throughput: these appear to be properties that emerge when the system can absorb change under control. They do not need to be pursued independently. They need to be designed in.

The firms that will matter most are not simply becoming better builders. They are learning to design the systems that builders operate in. Systems where the five implicit practises have become explicit, auditable, and machine-compatible.

What this means for consulting #

There is a natural temptation to respond to AI by scaling execution. More engineers. More tooling. Tighter integration with clients.

Those moves matter. But on their own, they do not resolve the new constraint. Without a system that can reliably absorb change, more output creates more downstream friction. This is the pattern already emerging at firms that have scaled engineering capacity without redesigning the delivery system underneath it.

Many firms will respond by scaling execution capacity. A smaller number will redesign how execution works. The difference will become visible quickly. Absorption, not throughput, is what determines whether AI-driven delivery compounds or collapses.

The opportunity is not to do more of what clients already struggle to absorb. It is to help clients design the system that makes absorption possible: the structure that carries intent into implementation, the control points where risk is bounded, the mechanisms that make behaviour reproducible and auditable.

That is where execution becomes scalable again.

The quiet shift #

Consulting is not disappearing. And it is not simply becoming engineering.

It is moving into a different layer of the enterprise: the design of how decisions become reality.

Most organisations are only starting to feel this pressure. Which means the window to define this space is still open.