In May 2026 the Financial Times ran an opinion column arguing that AI is forcing McKinsey and its peers to rethink how they price work — clients are questioning what advice is worth and are getting used to fees tied to finished outcomes. McKinsey has said as much on the record: Michael Birshan, the firm's managing partner for the UK, Ireland and Israel, has put roughly a quarter of global fees on outcomes-based pricing. Consultancy.uk asked the question more bluntly in January: is AI about to kill the billable hour?
My own version of it is cruder. Once a model can do in an afternoon what three analysts used to spend three weeks on, a timesheet stops being a credible way to describe value.
The pricing pressure isn't the interesting part, though. What it exposes is.
For most of a century the industry sold effort. A small number of expensive partners stood on a wide base of junior analysts doing research, building models, making slides. Effort scaled with headcount, headcount scaled with revenue, and it worked as long as the client couldn't see inside the box. Generative AI happens to be very good at exactly the work that filled the bottom of that pyramid, so the base is being priced down first, and firms are hunting for a margin model to replace it.
We built EGCA the other way round from the start, and the pricing argument had nothing to do with it. It came from something we kept noticing on site.
The asset that consultants keep leaving in their laptops
Every improvement engagement produces two things.
The first is the result: the saving, the extra tonnes, whatever the invoice was written against. That belongs to the client.
The second is the machinery that produced the result — the diagnostic logic, the benchmark set, the model, the report pack, the control routine that made a shift supervisor's decision visible by 10am instead of at month-end. Traditionally that machinery leaves site in the consultants' laptops. At the next client it gets rebuilt from scratch by different people, slightly worse or slightly better, and billed again.
The waste is in that rebuild, and in the fact that the client never keeps the part that would let them do it again without us.
Once you take that seriously, four things have to change.
1. Change who you hire
The traditional pyramid needs a wide base of graduates. Ours doesn't have one.
We hire senior - practitioners who have run or fixed operations rather than studied them - - and we slowly replacing the analyst layer with data scientists, machine learning engineers and a full in-house development team. That's a more expensive base and a smaller one, and it only makes sense if the technical people are building things that get used more than once.
It also changes what a consultant is for. The job isn't producing analysis by hand. It's knowing which analysis is worth producing, sitting with a plant manager while the answer lands badly, and making the change stick in a real operation with real constraints.
2. Turn the IP into software, not a folder
Every consulting firm says it has a knowledge base. Most have a shared drive.
A method sitting in a folder still has to be found, understood and rebuilt by whoever picks it up next. That usually costs about as much as starting from scratch, which is why templates get written once and used almost never.
We don't stop at writing the method down. When the same piece of work has been rebuilt by hand enough times, it goes to the development team and comes back as software — the diagnostic, the model, the control routine, running the same way on every site.
The client feels that as speed. Work that used to take the first few weeks of an engagement to construct now runs in days, so the first real answer arrives while there's still budget and appetite to act on it. It also changes what the client is paying for: our judgement about their operation, rather than our rebuild time.
3. Deliver past the roadmap
Conventional digital consulting produces a strategy, an architecture and an implementation roadmap, then hands the build to a systems integrator and the change to the client. The gap between those three parties is where most digital programmes die.
We build and implement. That comes with a deliberate limit on what we take on: we're not technical consultants and we don't take over our clients' engineering. What we build is the operating layer — the processes, tools, reports, modules, standard procedures — and then we coach the client's own team to run it. The test we apply is whether the thing still works six months after we've gone, when nobody is being paid to care about it.
4. Price the outcome
If you sell hours, every efficiency you find costs you revenue. That's a bad incentive to hand a client and an impossible position to defend once the client knows what the tooling can do.
We work to outcome-linked fees where the engagement allows it. The conversation is harder up front and easier at the end. Where the outcome can't be measured cleanly, or the client isn't willing to measure it, we bill for time and say so rather than dressing it up.
What it's actually produced
Three products so far, each one an engagement problem we'd solved by hand too many times.
SPINtex (spintex.ai) came out of short interval control — managing a shift while the shift is still running, rather than explaining it afterwards. It gives delivery teams a project status they can trust, with an AI assistant, Vera.
EntropyX (entropyx.co) came out of maintenance work. It's an industrial CMMS that starts from a 3D twin of the plant rather than a spreadsheet asset register, with a live maintenance view, shutdown and what-if planning, and an in-house AI agent, Shelby.
The Mine Maturity Assessment (Mine Maturity Assessment) came out of diagnostics. It's the structured version of the question every operation asks us in the first week: where do we actually stand against good, and which gap is worth closing first.
None of these replace the consulting work. They remove the part of it that was never worth paying a human to repeat.
What this doesn't mean
Software doesn't fix an operation. We've watched enough dashboards get installed over a broken planning process to be honest about that. A tool makes a good operating discipline cheaper to run and a bad one easier to see. It doesn't supply the discipline.
We also haven't stopped being a consulting firm. The judgement, the site time, the discussions with the client about what's actually true — that's still the product. What's changed is that the machinery behind the judgement can now leave with the client instead of with us or accelerate the client's pathway to the same answers that once took months.
EGCA Consulting works with mining and heavy industry across Africa and the Middle East on operational performance, cost, procurement, capital projects and operational software. If you want to see how any of this applies to your operation, book a diagnostic.
Sources: Financial Times, "How AI is forcing McKinsey and its peers to rethink pricing" (opinion, 23 May 2026); Business Insider, "AI is reshaping how McKinsey makes money" (Michael Birshan on outcomes-based pricing); Consultancy.uk, "Is AI about to kill the billable hour consulting model?" (January 2026).