Is your advisory business under threat from AI tools?

Ask a partner what a client is paying for and you get a good answer: judgement, pattern recognition across a hundred prior engagements, the nerve to say the thing the client does not want to hear. Ask what the client is actually invoiced for and the answer is different. It is hours. And a large share of those hours has always been the grind that sits underneath the judgement: reading the data room, cleaning the extract, pivoting the ledger, working out what really moved against plan, laying the finding out on a page a client will accept.

That grind is the part AI ate first. Not the judgement. The grind. Research that used to fill an analyst's day now takes an hour. A proposal that ate a weekend drafts itself to eighty percent. Trade coverage through 2026 has been blunt about the consequence: adoption across the profession now runs at roughly six firms in ten globally and higher still in the UK, delivery is measurably faster, and the firms applying AI across the whole delivery cycle report profitability several times that of the firms that have not. One of the largest firms has taken out around a tenth of its headcount across 2025 and 2026, concentrated in the junior research and synthesis roles.

So the threat is real, and it is widely misdiagnosed. The threat is not that your client replaces you with a chatbot. The threat is that your client runs the chatbot-shaped part of your invoice themselves, declines to pay for it, and still needs everything else you do. Your value did not fall. Your billable base did.

The margin problem in one sentence

If the work that used to need six people can now be done by two, and your price is a function of people, your revenue falls by two thirds even when you win every pitch.

That is the whole squeeze. It is not a demand problem. Demand for good advice is going nowhere, because a model that has never met the client, never sat through the steering meeting and never seen the politics in the room cannot tell anyone what to do on Monday. It is a pricing-basis problem. The basis was hours, and the hours are evaporating from the bottom of the pyramid upwards.

There are two responses. One is to shrink: cut the analyst bench, run leaner, accept the lower revenue and hope average deal size holds. Most firms are doing some version of this and it is a slow retreat, because that bench was also the training ground and the spare capacity that let you say yes to the next engagement. The other response is to keep the leverage and change what supplies it.

Deliver more work with fewer people, without dropping the standard

Leverage in consulting has always meant one senior person's judgement applied across many people's output. Nothing in that model requires those people to be human hires. It requires the output to be reliable enough that a partner will put their name on it. That is the bar, and it is precisely where most AI tools fail this job. A tool that writes a confident paragraph around a number it invented is not an analyst. It is a liability with good grammar. One fabricated figure in front of a client costs more than a year of subscription savings, which is why so much AI in advisory work stops at drafting the email and never reaches the work that actually carries the margin.

An AI that can genuinely take analyst work off the pyramid has four things to clear:

1. 1.It has to read the client's actual context. Not the internet's view of the client's sector. The management accounts, the tender responses, the CRM extract, the prior pack, the messy operational spreadsheet nobody has cleaned since March.
2. 2.It has to compute, not predict. The variance against plan is either the number in the data or it is not. A figure that merely looks correct is worthless in advisory work, because nobody downstream can tell the difference until it is too late.
3. 3.It has to structure an argument, not summarise text. A summary is not analysis. The client is buying a line of reasoning that runs from what the data says, to what it means, to what should happen next.
4. 4.It has to land in the format the work is delivered in. An answer trapped in a chat window still costs a consultant two hours to turn into something a client can be walked through.

Clear all four and the economics invert. The senior consultant stops being the bottleneck on production and goes back to being the bottleneck on judgement, which is the only bottleneck a client has ever paid a premium for.

What Nexlyr AI does that the tool you already have cannot

Nexlyr AI was built for that job specifically, and it is the reason the four points above are written the way they are. You hand it the engagement's real material: the spreadsheets, the reports, the proposals, the prior decks, the notes. It reads them. Where a figure is needed it writes the query and computes the answer on the actual data with a real analysis engine, so the number is derived rather than guessed, and any figure the source does not support is cut by code before it can reach a page. It then structures the finding the way a consultant would, choosing the framework that fits the situation rather than decorating a page with shapes, and it hands back a genuinely editable PowerPoint with a source trail on every slide. Not an image. Not a locked export. A file your team opens, argues with and changes.

It also remembers. The figures and the context of a client carry across engagements, kept separate per client, so the second piece of work starts from what it already knows about that business and lands sharper than the first. That is the compounding an analyst gives you after a year on an account, available on the first day.

Put next to what a firm typically has on the shelf today, the gap is not subtle. A general chat tool reads a file and predicts text: every figure comes back needing a manual check, and the output is still sitting in a chat window. A research platform searches external sources well and never touches the client's own numbers. An AI slide generator formats thinking you already did, which is the exact layer that just got commoditised, and invents the figures you then have to verify. Each of those does one part. None of them reads the client's real files, computes the figures on that data, builds the consulting argument and delivers the file you present.

The output is an eighty percent starting point, produced in minutes, on the client's own numbers. What the consultant adds is the twenty percent that was always the actual product: the read of the room, the thing the data cannot see and the recommendation they will stake their name on.

What that does to the arithmetic

Run this across a practice and the maths moves in the direction you want. The same senior team covers more engagements, because production is no longer what limits how many they can carry. The bench you kept spends its time on client-facing judgement instead of cleaning columns. Fixed-price and outcome-priced work, which is where clients are pushing everyone anyway, stops being a margin trap and becomes the best deal on your book, because your cost to deliver fell while the value of the decision did not.

And when a client asks why they are paying for a week of work an AI could do in an afternoon, you have a better answer than most of your competitors. They are not. They are paying for the afternoon, and for the twenty years of judgement that decided what the afternoon should be spent on.

Where the figures come from matters more than ever

There is a version of this future that goes badly for a firm, and it is not the one where AI gets better. It is the one where a firm adopts a plausible-sounding tool, a wrong number reaches a client, and the firm's reputation for rigour, which is the only thing it actually sells, takes the hit. Grounding is not a technical footnote in advisory work. It is the whole product. An AI analyst that computes every figure from the client's own files and prints nothing the source cannot support is the only kind you can safely put leverage on. Anything less and you have not added an analyst to the team. You have added a review burden that eats the time you thought you were saving.

The firms that come out ahead

The story in the trade press is that AI is coming for consulting. The truer version is narrower and more useful. AI is coming for the part of consulting that was always closer to production than to advice. The firms that lose are the ones whose revenue depended on selling that production by the hour and who answer by cutting until there is no leverage left. The firms that come out ahead keep the leverage, re-staff the bottom of the pyramid with analysts that never sleep, ground everything they produce in the client's real data and go back to selling what they were always meant to be selling.

Your advisory business is under threat from AI tools. The way through is to be the firm with more analysts than anyone else, at a cost that does not scale with headcount, and to be certain that every figure they produce is real.

Keep exploring

• Why AI tools invent numbers, and how source-grounding stops it
• From a raw spreadsheet to a finished analysis
• Market entry assessment