We surveyed leaders at 220 companies and asked two questions: Where do you think AI should be paying off, and where is it actually making or saving you money?
We’re talking about actual money: lower spending, hires you no longer need to make, or more gross profit. And that’s after paying for the AI, setting it up, checking its work, and fixing its mistakes. “It saves me a few hours” doesn’t count unless those hours change the numbers.
Put expectations next to results, and the gap gets pretty obvious.
Expected Returns vs. Actual Returns:
| Function | Actual ROI Rank | % Selecting | Expected ROI Rank | % Selecting |
|---|---|---|---|---|
| Engineering | 1 | 28% | 1 | 30% |
| Customer service | 2 | 23% | 4 | 12% |
| Sales and sales development | 3 | 17% | 3 | 18% |
| Customer success | 4 | 11% | 6 | 5% |
| Administrative operations | 5 | 8% | 5 | 7% |
| HR | 6 | 6% | 7 | 3% |
| Marketing | 7 | 5% | 2 | 23% |
| Finance / IT support | 8 | 2% | 8 | 2% |
Follow the Payroll
Engineering takes the top spot on both lists. You’ve got big teams, expensive people, and AI that can do a meaningful chunk of the work. But pumping out more code doesn’t pay the bills. Shipping the software you need with less contractor spend or fewer new hires does.
Customer service ranks second in actual returns but fourth in expectations. 23% picked it as their biggest financial win, compared with 12% who expected it to lead. The calls and tickets are already coming in, and you’re paying a team to handle them. If AI takes over enough of that work to lower your staffing or outsourcing bill without hurting service, the savings show up in the budget.
Sales and sales development hold third place on both lists. These teams spend a lot of time researching accounts, qualifying leads, chasing replies, and booking meetings. If AI takes over that work and delivers the same or better qualified pipeline at a lower cost, it pays off. Sending more emails doesn’t mean a damn thing if the economics don’t improve.
Customer success comes in fourth on actual returns, ahead of its sixth-place spot in expectations. Teams spend a lot of time getting customers onboarded, checking in, and keeping smaller accounts from falling through the cracks. If AI handles enough of that work to support more customers without adding CSMs, while retention and service hold up, the savings are real. More automated check-ins don’t count but avoiding the next hire you would otherwise need does.
Administrative operations sit fifth on both lists. Entering orders, processing documents, updating records, managing calendars, and booking travel aren’t flashy, but companies pay people to do them every day. If AI handles that work reliably and lowers the labor bill or avoids another hire, that’s real savings.
HR moves from seventh in expected ROI to sixth in actual returns. Screening candidates, scheduling interviews, handling onboarding paperwork, and answering routine employee questions all take paid time. If AI reduces agency fees or the staffing needed to handle that work, it pays off. Hiring volume and employee count determine how big the savings can get.
Marketing has the biggest gap. 23% expected it to deliver the biggest return. Only 5% picked it as their biggest actual financial win. AI pumps out posts, emails, ads, and decks, and suddenly everyone feels productive. But buyers don’t owe you money because you made more content. If acquisition costs haven’t dropped, agency spending hasn’t fallen, and gross profit hasn’t moved, you’ve built a content factory. Show me where it pays.
Finance and IT support sit eighth on both lists. AI can handle invoices, reconciliations, password resets, and routine tickets. But if you’ve only got one or two people doing that work, there’s only so much payroll to take out. A useful automation doesn’t automatically add up to a big company-wide return.
If the Numbers Don’t Change, Where’s the Return?
The question is pretty simple: how much work did AI take over, and what changed financially because of it?
Big teams, expensive labor, and repetitive work give you somewhere to start. But you still have to show the savings after paying for the AI and the people checking its work.
Everybody feels more productive. Great, where did the money go?
About the Research
The survey included leaders from 220 companies across software, business services, retail, healthcare, and financial services, ranging from 50 to 5,000 employees. Respondents included executives and department leaders responsible for budgets and AI deployments. Responses were collected during August and September 2026. Rankings reflect how often each function was named as delivering the largest annual net financial benefit. Financial claims were classified by supporting evidence: budget and spending records, hiring plans, operational metrics, or respondent estimates. Estimates were reported separately from documented results.