How AI Is Changing the Consulting Industry (and Why the Retainer Wins)

I've watched monthly-fee consulting firms produce more work with less effort while continuing to charge the same fee, and I've also watched clients cancel consulting contracts because they decided an AI chat could give them the advice they needed.

Both things can happen at once. AI can make a retainer more profitable for you and make the client less certain they need to keep paying it.

So if you are wondering how AI is changing the consulting industry, I would start with the agreement between you and the client. How you charge decides where a saving goes, but the reason the client keeps paying decides whether you get to keep that agreement at all.

The way I would use AI is to improve delivery and put some of the recovered time into useful work the client wasn't getting before. That work needs to answer a real problem, and someone who knows the client has to check it, otherwise you are just sending more material for them to deal with.

Let me explain the billing part first, because it is easy to miss when you are concentrating on making the team faster.

If you charge by the hour and the same task takes fewer billable hours, the invoice for that task comes down, assuming the rate and scope stay the same. The client receives the saving as a smaller bill.

You can still benefit, of course. If you have more work waiting, you can use the available hours on another project, so hourly billing doesn't mean AI is useless to you, it means the saving doesn't automatically stay in the fee for the first job.

With a fixed monthly retainer, the fee can stay where it is while the cost of delivering the agreed work falls. That leaves room for a better margin, provided you count review, rework and the tools as part of the cost to deliver.

But a fixed fee doesn't stay fixed forever. At renewal the client can ask for a lower price, more work, or an explanation of why the same fee is still worth paying, and you should be ready for that conversation.

Outcome-based pricing is another option, where payment depends on an agreed result rather than the hours spent. That creates its own work, because you and the client have to define the result, measure it and agree which parts you can actually control.

I wouldn't turn this into advice to move every client onto the same contract. A changing scope can make a fixed fee dangerous, and a result you cannot measure clearly can make an outcome agreement a long argument about who owes whom.

Start with the repeated work where you understand the cost and can describe what the client is buying. That is where the effect of the billing model becomes easier to see.

The big consulting firms are dealing with this too, but their starting point is different. Their traditional staffing model puts a lot of junior research, analysis and production under a smaller number of senior people, and those hours have supported the way they charge.

AI can now help with several of those tasks, so the model comes under pressure when the production hours fall. That does not mean AI does everything a junior consultant does, or that every reduction in consulting headcount was caused by AI.

Reporting on McKinsey described headcount falling by about 5,000 from its peak, while reporting on PwC put its global headcount reduction at 5,600 in its 2025 financial year. A headcount decline includes more than announced layoffs, and neither number isolates the effect of AI.

The pricing changes are more directly relevant to this piece. The Wall Street Journal reported that more than 30 percent of McKinsey's global fees were tied to client outcomes, and described the difficulty firms face in moving away from hourly work.

Bloomberg Tax reported managed-services targets of 20 to 25 percent of advisory revenue at PwC and 15 to 20 percent of US consulting fees at KPMG. Those were stated targets, not reported revenue shares already achieved.

If you run a smaller consulting firm, you may already work with a few senior people, a small bench and a junior supporting several projects. That is a shape I've seen, along with fixed monthly fees, so you don't necessarily have the same staffing model to unwind.

But having the right contract doesn't settle the client question. It just changes how much room you have to answer it.

In Source Global Research's review of its Q4 2024 client findings, 87 percent said they would pay more for consulting delivered with AI-enabled tools. In the same review, 58 percent expected consulting prices to fall.

Those are stated views, not a record of what every client later paid, and the figures don't tell us exactly how each respondent reconciled the two answers. My read is that clients can want a better result while questioning why they should still pay for the hours that result used to require.

The research also put client use of generative AI at 84 percent by Q4 2024. That tells you many buyers were already trying it themselves, not that demand for every kind of consulting was growing or guaranteed to hold.

And that is where the cancellations I've seen come in. The client engagement settles into a routine, the difficult early questions have been dealt with, and the consultant keeps doing work that no longer feels especially useful to the client.

Meanwhile the client is asking an AI chat their own questions and getting a clear answer quickly. It doesn't have to be as good as everything your firm can do, it only has to seem good enough for the work they think they are paying you for now.

From your side, you may be maintaining the relationship and making sure the agreed deliverables arrive. From their side, the same monthly bill has started paying for something they believe they can get elsewhere with less effort.

I would not wait for the cancellation email to have that conversation. Ask what the client is trying to solve next, what has become harder in their business, and whether the work you send helps them make those decisions.

That is different from asking if they liked the report. A report can be well written and arrive on time while answering a question that no longer matters very much.

The firms I've watched hold their clients have used AI to deliver useful extras beyond the original scope, market research, analysis of the client's own data, competitor work, social strategy ideas, reading and content suggestions.

These were produced with AI and checked by a person. The lower cost made that extra work possible, and the person supplied the understanding of which part would help this client rather than just fill another document.

I don't have a measured retention lift to give you for that pattern. What I have seen is firms doing more relevant work for their clients at a cost they could carry, rather than treating every hour recovered as a reason to reduce the relationship to the contract minimum.

And please don't read that as a promise to do unlimited work for a fixed fee. You still need scope, you still need to know what it costs, and an extra that becomes a standing expectation may belong in the next agreement.

The practical move is to keep a short list of the problems you hear from each client. Build it from conversations, the numbers they share and what is changing in their business, then decide which problem deserves attention before it becomes urgent.

If a client is worried about losing a particular kind of customer, a focused piece of analysis on that question could help. A general report on trends in their industry may look more impressive and tell them much less about what to do.

That example is the distinction I want you to make. Choose the question from the client's situation before you ask an agent to produce the work, otherwise the tool's ability to generate material starts deciding what you send.

I would begin with one useful extra in the month, not a new weekly reporting obligation. That is a starting recommendation, not a cadence I can prove is right for every relationship, and the client may prefer a short conversation to another file.

The cost needs checking too. AI production is cheaper for many of these tasks, but someone has to prepare the input, run the work, verify it and decide what to do if the answer is poor.

Count that whole job before saying the retainer has become more profitable. The model and its settings can change the bill, and the amount of senior review can change it more, so measure the setup you actually use rather than assuming every extra is nearly free.

This is the consulting version of giving saved time a purpose. The hours don't become retention work by themselves, the manager has to make room for it and someone has to own the client question.

There is a very clear counterexample. If you send a monthly bundle of generic AI reports, the client has to work out what is relevant and whether any of it is true.

You have given them another job, and you have also shown them something they could generate in their own chat window. More output can make the substitution easier when the judgment is missing.

So every extra needs a named person responsible for checking it. That person needs to understand the client, not only whether the sentences read well.

They should check the source of a number, whether the analysis fits the client's data, whether a recommendation conflicts with something the client has already decided, and whether the firm is making a claim it can support. That checking time belongs in the plan before you offer the extra.

In October 2025, Deloitte Australia agreed to a partial refund on an A$440,000 government report after errors including nonexistent references and citations were identified. The revised report disclosed use of Azure OpenAI GPT-4o.

That is an Australian-dollar contract amount, not US dollars, and the partial refund was not a refund of the whole fee. The useful warning is that a large firm's name and a polished report did not keep unsupported references out of the delivered work.

Your extra carries your name too. Calling it a bonus doesn't make the client responsible for checking it, and a deadline doesn't turn a generated citation into a source.

So will AI replace consultants? It has already replaced some consulting relationships in the cases I've seen, and it can take over parts of research and production, but neither fact settles what happens to your firm.

The safer question is whether the client still receives something useful from your judgment and your knowledge of their business. If they do, AI can help you deliver more of it, if they don't, changing the invoice from hours to a retainer won't fix the missing reason to renew.

For the next client review, bring the current agreement, what delivery actually costs, and the short list of problems the client is facing. Pick one piece of work you can improve or add, decide who will check it, and make sure the client wants the result before you produce it.