Most consulting firms are spending their AI budgets on delivery: faster analysis, leaner teams, quicker reports. That investment is necessary, and competitors will largely match it, so the margin it buys will not last. The firms that pull ahead will use AI to find work no client has yet asked for.

Work a client can specify is becoming a commodity

A brief is evidence that the hard part is done. The client has noticed the problem, decided it is worth paying for and described it well enough for several firms to bid. What remains is execution, which is where AI is strongest.

On some assignments, a small team can now cover ground that once needed a much larger one. Clients' own staff can get much of the way with the same tools. Firms that lacked the specialist depth to compete can now write credible proposals. More capacity chasing the same briefs has a predictable effect on price.

Recent economic research describes this as a shift from scarce solutions to scarce problems. When solving becomes cheap, the constraint is knowing what is worth solving.

AI opens up problems that were never worth looking for

Every organisation carries problems that are real but too small or scattered to investigate one at a time. They sit well below the agenda a senior partner discusses with a chief executive, and until now the economics of consulting have left them alone. That is changing.

• A reward consultancy can test pay decisions across every role in a business, not a sample, and see where equal-pay exposure is building.
• A procurement adviser can see how the same item is specified differently across forty sites, at different prices.
• A regulatory specialist can read every customer complaint and spot the patterns a regulator would see, before the regulator does.
• A pensions adviser can find the data errors in a scheme that will distort an insurer's buy-out price.

Each of these is a finding before it is an assignment. The firm that sees it first gets to define the work, instead of bidding for it. This is ground where specialist firms have the advantage, because the largest firms' cost base is built for large assignments.

A client with its own data and the same tools could, in principle, find these problems itself. In practice it lacks the comparison. A pensions adviser that has seen two hundred schemes knows what normal looks like, and can tell when one scheme's data is out of line. A single client sees only its own. The defensible asset is that knowledge across clients, built from aggregated, anonymised patterns under rights agreed at the start, and few clients can build it for themselves. Whether a firm may keep and reuse that knowledge is a question of contract, which we take up in "Your best AI work may be giving the firm away".

Finding the problem is only half the job

A list of anomalies is not a consulting opportunity. Somebody with authority has to care enough to act. Firms that send clients long lists of things they could fix will be thanked and ignored.

The value goes to a firm that combines three things: ongoing access to the client's data and the comparisons to interpret it, the judgement to tell a material finding from a merely interesting one, and the standing with a decision-maker to turn it into a funded decision. A firm with only the data, or only the relationship, will end up feeding work to someone else. A long relationship with the chief executive is valuable, but it will not reveal a problem buried in the client's data.

This changes how consulting is sold. Business development and delivery start to merge. The most valuable people become those who can look at evidence and say, with authority, "this matters, and here is what to do about it". As the cost of looking falls, many firms will give discovery away to win the work, and expect to be paid on what the finding is worth. Gain-share and outcome-based fees, long treated with suspicion, will return as a result.

What this means for investors and leaders

For an investor, historic delivery margin shows how well a firm served demand in the old conditions. It says little about whether the firm can create demand in the new ones. We would look at the share of revenue from work the firm originated rather than won in competition, how much client data it can see on an ongoing basis, and whether its senior people are known for finding problems or for fixing them. Most firms do not track the first of these. If a firm cannot say what share of its work it originated, that is itself an answer. A strong record in competitive tenders shows that a firm is good at answering briefs, not that it can write them.

For leaders, the question is where the AI budget goes. Some of it belongs in delivery. A deliberate share belongs in discovery: the tools, data access and commercial terms that let the firm see problems first. The right balance depends on where the firm's advantage lies, and it should be decided deliberately.