Nearly 750 CFOs, One Direct Answer

A March 2026 working paper from the Federal Reserve Bank of Atlanta surveyed nearly 750 chief financial officers, the executives who approve AI budgets and headcount decisions. When asked why they invest in AI, executives rated reducing labor costs among the least important of nine reasons, scoring it 2.0 out of 4, well behind improving production efficiency at 2.9 and improving decision making at 2.7.

The strongest productivity gains showed up somewhere else entirely: developing better products and reaching or serving customers more effectively. Put simply, the bigger opportunity is not cutting the pie into fewer pieces. It is making the pie bigger.

Same Team, Bigger Output

Consider a software development agency deciding how to direct AI toward its coding work.

The cost cutting path: AI performs the routine coding. Half the development staff is released. The employees who remain spend months wondering whose role is next, and morale and retention suffer along with it. The team delivers the same number of applications it always did, at a reduced payroll, and profit increases this quarter.

The growth path: AI performs the routine coding. The team stays intact, and sees AI as something that expands their work rather than threatens it. Freed from repetitive tasks, developers who once shipped five prototypes a month now ship fifteen. They test more ideas and reach markets that were out of reach before.

Why the Payoff Takes Longer

This also explains what we previously called The AI Productivity Paradox. A layoff shows up in next quarter’s numbers immediately. A team using AI to build new products or reach new customers needs time before that value reaches the income statement, and the researchers point to this delay as the likely reason perceived productivity gains consistently outpace what shows up in revenue.

Large firms are more likely to expect AI driven job cuts. Smaller firms expect modest job gains instead. That points to two different uses of AI: consolidation where firms already have scale, and expansion where they still lack capacity.

Redesign the Role, Do Not Just Eliminate It

The report found little evidence of major job losses from AI so far, projecting the total effect at under half a percent of employment in 2026. But the type of work is shifting underneath that stable number. Routine clerical work is expected to shrink, and skilled technical roles are expected to grow in relative importance over the next three years.

That does not mean every role stays exactly as it is. It means the choice is not only between unchanged and eliminated. Redeploy before you reduce: before cutting a role, ask whether that capacity can move toward customer engagement, product development, or the technical work AI makes more valuable.

Start With the Outcome

Decide what the AI investment is actually for before you make it. Some processes are exactly right for automation and lower cost, and saying so plainly is fine. What does not hold up is assuming cost reduction is the highest value use of AI without checking, the distinction covered in Start With the Work, Not the AI.

Three questions are worth asking for any AI initiative. What work can AI do better or faster? What capacity does that free up for our people? Where can that capacity go to improve products, serve customers, or generate new revenue? Measure success against that answer, not against how much work disappeared.