The part of consulting AI still can’t do

This summer the Financial Times asked a question most of us in consulting would rather answer on our own terms: who needs consultants in the age of AI? If you haven’t read it, suffice to say Robert Armstrong’s answer wasn’t a comfortable one for anyone still billing by the hour:

AI is genuinely faster and cheaper at a large share of what consulting has charged for. Market research, benchmarking, structuring a problem, building the deck that walks a board through eighteen months of analysis (a model does most of that today, in minutes, at a cost that makes the old day-rate look strange). Pretending otherwise is the fastest way to lose this argument. Work that used to take a graduate analyst a week now takes a well-prompted model in an afternoon, and clients already know it. Some of them are doing it themselves before they even call us.

Armstrong’s framing of it stuck with me: for years the analysis was the meal and human judgement was the seasoning. AI is eating the meal. What’s left, as he puts it, is “just sauce.” That’s not a comfortable image, and I’m not going to argue it’s wrong.

Judgement is the part that doesn’t automate

But sauce is not nothing, and it was never the byproduct. The part of consulting that doesn’t automate is the part that was never really about the deck. Getting a room of executives with different incentives and different readings of the situation to agree on a decision. Two people can look at exactly the same model output and reach opposite conclusions about what to do with it. This is where the actual work of consulting happens.

Building the trust that makes someone act on a recommendation instead of shelving it after the meeting ends, taking into account the politics that never once show up in a slide. That’s judgement, and it can’t be assembled from data. It’s built from having sat in enough of those rooms, made enough calls, and lived with the consequences of what happened next.

Paying for results, not effort

This changes the question a client should be asking a consultant. Instead of questioning how many hours were spent on a consultation, they should ask “what changed because you were there?”

Picture two engagements side by side. One consultant delivers a forty-page deck and a day-rate invoice, and six months later nobody can say with confidence what actually happened as a result. Another sits in the room while the recommendation gets contested, works the politics it exposes, and is still involved when the plan either holds or doesn’t. Only one of those is being paid for a result. The other is being paid for effort.

An hourly engagement buys presence and effort, but an outcome-based engagement means buying a result, whether it’s a restructure that shipped, a strategy that survived, a change that stuck even after the consultant left. Outcomes are harder to price and harder to promise, but they are also much harder to fake. And let’s face it: it’s the only version of the business model that AI doesn’t chip away at every time the tools get better.

Will AI change the profession? It already has. That’s not the issue here. It’s whether or not you’re pricing and building your own practice in a way that accommodates that shift, or still selling your time out of habit because that’s the way it’s always been done.

Built around judgement

At The Principals Group, we built our infrastructure around that answer: the research, the benchmarking, the first-draft thinking runs on AI, so our Principals spend less time producing the meal and more time on the part no model does. That means aligning the room, holding the trust, seeing the change through. The tools handle the intelligence work, but our Principals handle the judgement.

The obituary for consulting has been written before, and it was wrong before. AI really is taking the meal...that part is true. So here’s the question I’d leave you with, the same one Armstrong asks: of everything you currently bill for, how much of it is sauce?