Skip to content
Skip to content

Where AI creates operating leverage; and where it doesn't

Where AI creates operating leverage; and where it doesn't

Where AI creates operating leverage; and where it doesn't

Where AI creates operating leverage; and where it doesn't

Headshot of Andrew Jin

Andrew Jin

Andrew Jin

Share

The tool gets the credit, but the task does the work

Every quarter brings a new claim that some platform, product, or assistant is the reason a firm pulled ahead. The evidence rarely supports it. When operating leverage shows up in knowledge work, it shows up in the same places regardless of which vendor's logo is on the login screen, and it stays absent in the same places no matter how much is spent. The variable that moves is the work, not the purchase.

We looked at the public task structure of professional-services work to find out what actually predicts where AI creates leverage. Using O*NET data for management analysts, whom the Bureau of Labor Statistics describes as often called management consultants, together with economy-wide adoption figures from the Census Bureau, we found a common pattern. Leverage follows the shape of the task, not the tool applied to it. We call this the task-shape rule.

A consultant's job is almost entirely structured knowledge work

O*NET rates every work activity in an occupation for importance on a 0 to 100 scale. For management analysts, the highest-rated activity is getting information, at 93, and nothing physical comes near it. Inspecting equipment sits at 25, operating vehicles and machinery at 14 to 17, and repairing equipment at 6 to 8. The work that a specific machine has historically defined is the least important thing a consultant does.

That gap is the exhibit below, and it is the opening half of the task-shape rule. The occupation is built out of getting, analyzing, and interpreting information, and it is almost entirely detached from the physical operations where the choice of equipment once decided the outcome. When the job is information end to end, the leverage has to come from doing information work faster and more consistently.

For a management consultant, the 10 information and analysis tasks rate 63 to 93 of 100 in importance, led by getting information at 93, while the physical and equipment tasks a specific machine touches rank last, from 25 down to 6.

Source: O*NET OnLine, Management Analysts (13-1111.00), Work Activities importance, accessed September 2026.

Leverage compounds on the repeated, checkable core

10 of the occupation's 41 work activities are information handling and analysis, and they cluster near the top: getting information at 93, analyzing data and interpreting it for others at 85 each, working with computers at 82, processing information at 79, down to estimating quantities at 63. The set averages 79 of 100. This is the repeated, structured core of the job, and it is where automation compounds across engagements rather than helping a single time.

By contrast, the 8 physical and equipment activities in the same occupation average under 15 of 100, a gap of roughly 5 times below the information core. No software choice closes that gap, because the gap is a property of the job, not of the tool. The leverage lives in the top band and is missing from the bottom band, whatever a firm buys.

The core is also exacting, which is what makes it automatable in a defensible way. Among the people who do the work, 85 percent call being highly accurate very or extremely important, and all of them report using e-mail every day. Exact, checkable, text-and-data work performed the same way across many cases is the definition of a task where a repeated process pays off. That is the task-shape rule in a single line: repeated plus structured equals leverage.

The tool is the least predictive variable

If leverage came from the tool, adoption would cluster around whichever product won the last news cycle. It does not. It clusters around the information intensity of the work. When U.S. firms began reporting AI use in late 2023, only 3.8 percent used it economy-wide, but the information sector was already at 13.8 percent and professional, scientific, and technical services at 9.1 percent, the 2 most information-heavy corners of the economy leading every other.

The government's own instrument makes the point most plainly. The Census Bureau's current survey measures AI use across 15 distinct business functions, from finance to human resources to customer service to research. AI is not a single tool doing a narrow job.

A firm can run it in finance and in marketing and in research at the same time, 3 different tasks with 3 different shapes, which makes the tool a shared input rather than the cause of any single result. What predicts where it lands is the structure of the task in each function, not the brand of the tool, which is the closing half of the task-shape rule.

Adoption rises with information intensity and with scale

2.5 years after that 2023 reading, the national adoption rate had roughly quintupled, from 3.8 percent to 19.8 percent, a 5.2 times increase. The information sector itself climbed from 13.8 percent to 39.7 percent, nearly tripling in level while holding its place at the top. Alongside it, nearly 40 percent of information-sector firms and about 34 percent of finance firms used AI, against roughly 20 percent economy-wide. The ranking barely moved even as the absolute numbers climbed, because the ranking was never about the tools.

Scale points the same way. In the most recent data, 37 percent of firms with at least 250 employees used AI, and 32 percent of firms with 100 to 249 employees, both well above the economy-wide rate. Adoption rises with size, exactly as it should if the payoff comes from applying a structured process across a larger book of repeated work rather than from any single purchase.

3 questions decide where the leverage is

The task-shape rule turns a shopping problem into a diagnostic problem. Instead of asking which tool to buy, a firm gets further by asking which of its own tasks will reward automation at all, and then pointing whatever it buys at those. 3 questions sort the work, in this order.

  • Is the task repeated across cases? Leverage compounds only when the same work recurs. A process run annually returns almost nothing on automation, while the analysis a team performs on every engagement returns it many times over.

  • Is the task structured and checkable? The core consulting activities score 79 of 100 on importance and draw an 85 percent vote for exactness. Work with a right answer and a repeatable method is where a machine can be trusted and audited.

  • Is the task information work, not physical? The activities that a specific device would touch rank last, at 6 to 25 of 100. The further a task sits from getting and analyzing information, the less any tool will move it.

The question is not which tool but which tasks

Management consulting is not a shrinking profession waiting to be replaced. It is about 1.1 million U.S. jobs at a median wage above $100,000, and the Bureau of Labor Statistics projects it to grow 10 percent through 2035, adding roughly 109,200 positions, much faster than the average occupation. The work being reshaped is expanding, which means the leverage is a growing prize rather than a passing cut, and it accrues to whoever organizes the repeated core ahead of rivals.

The firms that capture it will be the ones that stop debating tools and start sorting tasks. The task-shape rule says the answer was never in the software. It was in the work, in whichever parts of it are repeated, structured, and made of information. Buy any capable tool a firm likes; the leverage was decided before anyone opened it, by the shape of the task it was pointed at.

Related posts

Related posts

Palantir alternatives

Best Palantir alternatives for mid-market companies in 2026

Best Palantir alternatives for mid-market companies in 2026

Eight Palantir alternatives for companies of 100 to 2,000 people and PE portfolios: what each one is, who it fits, what it publishes about price, and where it falls short.

Eight Palantir alternatives for companies of 100 to 2,000 people and PE portfolios: what each one is, who it fits, what it publishes about price, and where it falls short.

Diligence before IC

Commercial diligence before IC: What converges in 30 days and what never will

Commercial diligence before IC: What converges in 30 days and what never will

Sponsor buyers set their own decision window at about a month, so the work that matters before IC is classifying the thesis variables rather than collecting more of them.

Sponsor buyers set their own decision window at about a month, so the work that matters before IC is classifying the thesis variables rather than collecting more of them.

Thesis to execution

From investment thesis to execution: The variables to re-measure after close

From investment thesis to execution: The variables to re-measure after close

2 operating variables carry 93 percent of the unlevered gain in the current-cycle attribution of buyout value creation, yet after close the scoreboard becomes adjusted EBITDA.

2 operating variables carry 93 percent of the unlevered gain in the current-cycle attribution of buyout value creation, yet after close the scoreboard becomes adjusted EBITDA.

Start with the decision in front of you.

Tell us what your team needs to understand, evaluate, or deliver.