Positioning the Practice When the Deliverable Gets Cheap

Consulting has a sharper version of the disclosure question than most professions, because the client is paying for expert labour and has a direct interest in how it was produced.

The range of current practice runs from full disclosure to silence, and most firms have not decided deliberately, which means individual consultants are improvising inconsistently.

Three positions, and the case for each.

Disclose the use openly, describing where AI is used and where human judgement is applied. Strongest position, increasingly expected, and it turns a potential objection into a demonstration of rigour if the sourcing discipline is real.

Disclose on request. Workable, and it means the first time it comes up is in a conversation you did not choose.

Address it in the engagement letter. The most robust, since it establishes the position before any work happens and gives the client the chance to set constraints.

What is not viable is silence combined with a client who assumes otherwise. The asymmetry from the marketing cursus holds and is sharper here: a client who discovers that a deliverable they paid partner rates for was substantially generated will not weigh whether disclosure was required.

And the practical point that resolves most of it. If you would be uncomfortable telling the client, that discomfort is information about the arrangement rather than about the disclosure.

A large share of current consulting demand is advising clients on AI, which creates a specific credibility position worth thinking about deliberately.

The opportunity is real. Organisations need help with exactly the material in this catalogue: what to adopt, how to govern it, whether the vendor claims hold, how to measure whether it worked. Consultants are well placed because the work is cross-functional and clients lack the internal expertise.

The credibility risk is equally real, and it has two forms.

Advising on something you have not done. A firm recommending an AI governance framework it has not implemented internally is in a weak position the moment a client asks how they did it. The question is increasingly asked.

And advising on rigour you are not applying. A consultant warning a client about hallucination risk, in a deck containing an unverified statistic, has a problem that is not merely embarrassing.

The position that works is straightforward. Do it internally first, honestly, including the parts that did not work. A firm that can say we implemented this, here is what we learned, here is what we got wrong is considerably more useful and more credible than one presenting a framework abstractly.

And be honest about the evidence base. Much of what is sold as AI strategy rests on claims that do not survive scrutiny, and the AI transformation evidence lessons in this catalogue make that point. A consultant repeating unsourced productivity figures to a client is doing the thing the client hired them to see through.

Sorting a consulting proposition by how exposed it is.

Most exposed: information asymmetry. Practices whose value was knowing things the client could not easily find out. That asymmetry has narrowed sharply, since a client with a good model and an afternoon can now produce the landscape summary that used to be a deliverable.

Exposed: production capacity. Practices selling the ability to produce large volumes of analysis and documentation. The volume is now cheap and the differentiation was never the volume.

Defensible: pattern recognition across engagements. Having seen this problem in eleven organisations, and knowing which of the plausible answers fails in practice. A model has read about the problem; it has not watched three implementations fail for reasons nobody wrote down.

Defensible: execution knowledge. Knowing what an organisation of this shape can actually absorb, which is the difference between a correct recommendation and a useful one.

And durable: accountability and cover. Someone external, with a reputation at stake, willing to say the thing and be held to it. That value has nothing to do with labour and is unaffected.

The strategic reading: the top two were always the least defensible parts of the proposition, and the compression exposed rather than created that.

The compression affects firm sizes differently, and the independent consultant's position is stronger than it first appears.

What large firms had that small ones did not. Leverage: the ability to put six juniors on a research task. Reach: offices and sector coverage. And process: methodology, quality assurance, the machinery of a large engagement.

The first of those has narrowed substantially. An independent consultant with good tooling can now produce the research and documentation output that previously required a team, which removes a real barrier to competing for certain work.

What that does not remove. Reach, accountability at scale, the ability to absorb a large engagement's risk, and the brand that provides cover. A board wanting an external endorsement may specifically want a recognisable name, and that requirement is unaffected by efficiency.

So the sensible independent positioning sharpens rather than broadens. Deep expertise in a specific domain, where pattern recognition across engagements is the product, delivered by the person the client actually wanted. That was always the independent's strength and the production disadvantage partly obscured it.

And for a small firm the pricing question resolves more easily, because there is no pyramid to protect. A firm without leveraged junior time to defend can move to value pricing more readily than one whose economics depend on billing hours it no longer needs.

The uncomfortable corollary: the pressure lands hardest on mid-sized firms selling capacity rather than distinctive expertise.

The question that is arriving and that deserves a prepared answer rather than an improvised one.

What clients can now genuinely do without you. Landscape and competitor summaries. First-pass literature reviews. Drafting their own strategy documents. Much of the descriptive analysis that used to justify an engagement.

That is a real capability shift and pretending otherwise is unwise, because the client will test it.

What they still cannot do, and this is the substantive answer.

They cannot see their own organisation clearly. Internal analysis is constrained by what people can say to each other, and the value of an external party is partly that people tell them things they will not tell a colleague.

They do not have the comparison set. Knowing that your integration timeline is twice what comparable organisations achieve requires having seen the comparable organisations.

They cannot provide their own cover. A recommendation from the internal team competes with other internal views; the same recommendation from outside settles it, which is a political fact rather than an intellectual one.

And they lack the time. Capability is not capacity, and a client team with the ability to do the analysis frequently does not have the weeks.

The positioning that follows. Sell the things a client structurally cannot do for themselves, and be relaxed about the things they now can. A firm defending the research deliverable is defending the wrong ground.

Firms have review processes, and generated content changes what those processes need to catch.

What traditional review catches. Analytical errors, weak argument, poor structure, inconsistency with the firm's position, tone. All still necessary.

What it does not catch, because it was never designed to. A fabricated statistic with a plausible source attached. A causal claim the data does not support, expressed fluently. Analysis that is conventional rather than insightful, which reads as competent. And a commitment in a proposal that nobody agreed.

So the review needs a distinct pass, in the same spirit as the two-pass content review in the marketing cursus. One pass for the argument, which is what reviewers already do. One pass for provenance, checking that each number traces to a source someone opened and that judgement is labelled as judgement.

The second pass is boring, fast, and it is the one that prevents the incident.

Two further notes for a firm. The reviewer should ask whether the analysis says anything non-obvious, which is the check against the synthesis trap and which a partner is well placed to make. And a rising proportion of deliverables that read well and say little is a leading indicator worth watching, because it is what convention-at-scale looks like from the inside.

And whoever signs the deliverable owns it. The AI produced it is not available as an explanation to a client, a regulator or a court, and stating that internally before it is tested is worth doing.

A firm adopting this should be able to say what changed, and the available metrics need care for the same reason they do in marketing.

What is easy and misleading. Hours saved per deliverable. Real, and it measures the input that got cheaper rather than any outcome. A firm reporting this alone is reporting that its cost base fell, which is a margin statement rather than a quality one.

What is worth measuring.

Where the recovered time went. Into more engagements, into more depth per engagement, or into margin. All three are legitimate and they are different strategies, and only tracking it reveals which one the firm actually chose.

Win rate on proposals, watched carefully. If proposal volume rose and win rate fell, the differentiation problem is operating and the efficiency is costing revenue.

Deliverable quality proxies: client-raised corrections, questions the deliverable should have answered, and repeat engagement rate.

And provenance failures caught in review, which is the counterweight metric. A rising count means the sourcing discipline is not holding, and it is much better found internally.

The honest position, and it mirrors the marketing conclusion. A consulting firm cannot cleanly attribute client outcomes to its tooling, so measure the process change, watch the win rate and the correction rate, and be suspicious of a productivity claim that arrived without either.

What this cursus adds up to for the profession.

The compression is real and uneven. Research and production compress substantially. Analysis compresses where the data is clean, which is rarer than assumed. Synthesis compresses into convention, which reads like insight and is not. Judgement and cover do not compress at all.

The pricing model is exposed. Billing inputs for work whose inputs collapsed is unstable, and the profession has deferred the value-pricing conversation for a long time. The most defensible response is to deliver more depth for the same fee, which converts efficiency into quality rather than margin, and it is the least discussed internally.

The development model is exposed further. The junior work that built consulting judgement is exactly what compressed, the pyramid economics depended on that work being billable, and nobody currently has a convincing answer to what replaces it.

The defensible proposition is narrower and clearer than before. Pattern recognition across engagements, execution knowledge about what an organisation can absorb, seeing the client more clearly than they see themselves, and accountability. None of that is threatened, and a firm that could not previously articulate its value in those terms now has to.

And the professional discipline that matters most is unglamorous. Every number traces to a source someone opened, judgement is labelled as judgement, and whoever signs it owns it. That was always true. What changed is that it is now much easier to produce something persuasive that violates it.