AI Talk Nearly Doubled Outside Tech. Proof of Payoff Is Rare

Summary

Executives at S&P 500 companies outside technology now talk about AI on most of their earnings calls. From July to September 2026 they did on 60.5% of calls, up from 32.1% in January to March 2024.

Few of them put a number on what it delivers. Last quarter an executive attached a figure to what AI saved, sped up or earned on 25 of 387 calls, 6.5%. That is up from 1.6% of calls in 2024, but it means that on about nine in ten calls where executives talked about AI, they gave no number for its payoff.

The numbers come from a small group. 86 companies have put a figure on AI at least once since January 2024, and nine of them, led by Airbnb and UnitedHealth, account for 47 of the 158 calls. The most common figure is a savings target for a company-wide cost program in which AI is one lever among several. The rest are narrow results in customer service, claims and document processing, and software development.

Analysts have started to ask. An analyst asked about AI savings, productivity or returns on 6.7% of calls last quarter, against about 1% in 2024, and on 15.1% of the first 86 calls of October.

Abstract

We read the earnings calls of the companies in the S&P 500 outside the technology sector, 4,335 calls by 403 companies held from 1 January 2024 to 23 October 2026, using the speaker-by-speaker transcripts in the Equibles earnings call archive. We count a call as mentioning AI when an executive used the term, and as putting a number on AI when an executive attached a specific figure to what AI delivered: money saved, time or cost cut, a share of work automated, productivity, or revenue from AI products, achieved or targeted. Every candidate statement was read by hand.

The share of calls where executives mentioned AI rose from 32.1% in January to March 2024 to 60.5% in July to September 2026. The share where they put a number on it rose from 1.8% to 6.5%, or from 5.6% to 10.7% of the calls that mentioned AI. Financial services companies did both most often, mentioning AI on 89.1% of their calls in January to September 2026 and quantifying it on 11.4%. Energy and basic materials companies mentioned AI on 50% and 42% of their calls and never put a number on it. Of the 70 calls with a number in January to September 2026, 23 were company-wide savings or productivity programs, 17 back-office work such as claims and documents, 12 AI products and AI-assisted sales, 8 customer service, 7 software development and 3 product development.

Twice the talk, a few more numbers

In early 2024 executives mentioned AI on about a third of non-tech calls. The share passed half in July to September 2025 and has held at about 60% for three quarters.

Calls where an executive put a figure on what AI delivered grew faster, from a low base. They went from 1.8% of calls in January to March 2024 to 6.5% in July to September 2026, with the steepest rise in the second half of 2025.

October covers the first 86 calls of the reporting season, mostly banks and other financial companies, so its share is not comparable with a full quarter.

Early examples were rare and specific. On Walmart (WMT $120.29 +0.33%)'s August 2024 call, the chief executive said the company had used large language models to create or improve over 850 million pieces of data in its product catalog.

“Without the use of generative AI, this work would have required nearly 100 times the current headcount to complete in the same amount of time.”

Doug McMillon, chief executive, Walmart, earnings call, 15 August 2024

Who puts a number on it

In January to September 2026, executives at 316 of the 399 non-tech companies in the sample talked about AI on at least one call. Executives at 49 of them put a number on what it delivered.

The same companies come back quarter after quarter. Airbnb (ABNB $171.05 -0.56%) and UnitedHealth (UNH $410.70 -1.77%) gave a figure on six of their calls since January 2024, and Nasdaq (NDAQ $104.72 +2.11%), C.H. Robinson (CHRW $140.15 -0.72%), Expedia (EXPE $283.97 -1.74%), FactSet (FDS $323.39 -2.39%), Global Payments (GPN $81.86 -1.16%), Equifax (EFX $178.56 -2.08%) and WTW (WTW $324.67 -0.10%) on five each. Those nine companies account for 47 of the 158 calls with a number.

Financial services companies, banks, insurers, exchanges, card networks and data providers, talk about AI on almost every call and quantify it most often. Energy and basic materials companies talked about AI on 50% and 42% of their calls, often as a source of demand for power, gas or materials, but none put a number on what AI did for their own operations.

Where the numbers come from

The largest group is a savings or productivity target for the whole company, with AI named as a driver. WTW tied a new program directly to AI on its July 2026 call.

“With the latest significant AI advancements, we intend to create a step change in performance. We expect to capture efficiencies and generate approximately $400 million in run rate savings through an investment of approximately $625 million, reflecting a disciplined cash cost to achieve ratio of about 1.6.”

Carl Hess, chief executive, WTW, earnings call, 30 July 2026

Equifax doubled its target on its July 2026 call, and Sysco (SYY $82.16 +0.97%) set one on its August call.

“The pace of AI adoption inside Equifax is accelerating rapidly, which allows us to double our AI for EFX productivity goal from $75 million to $150 million from 2026 to 2028.”

Mark Begor, chief executive, Equifax, earnings call, 21 July 2026

“We expect the combined benefit of these efforts to deliver approximately 100 million of in-year savings in fiscal 2027, inclusive of the cost out savings we shared on our Q3 earnings call.”

Kevin Hourican, chief executive, Sysco, earnings call, 4 August 2026, on AI growth and efficiency projects

Many of these targets credit AI alongside other levers, such as reorganizations, procurement or offshoring, so the figure is not AI's alone. Of the 25 calls with a number last quarter, 19 credited the figure to AI alone, and 6 named AI as one of several drivers.

The narrower results are about time. GE Aerospace (GE $288.43 +0.98%) described its supply chain on its July 2026 call, and CVS Health (CVS $92.53 -1.01%) its claims on its August call.

“By deploying Flight Deck to simplify and then using AI to automate the process, we cut the number of demand signals in half and reduce processing time by nearly 90% across 190 parts.”

Larry Culp, chief executive, GE Aerospace, earnings call, 16 July 2026

“We've recently launched a new AI-enabled claims assist manager, which will reduce our processing time by over 20% and accelerate payment for providers on hundreds of millions of claims every year.”

David Joyner, chief executive, CVS Health, earnings call, 5 August 2026

Customer service gives the most concrete figures. Airbnb has given a figure for its AI customer service agent on every call since May 2025, and on its August 2026 call it tied a cost figure to it.

“Nearly 45% of issues that start with our AI assistant are now resolved without a human agent.”

Brian Chesky, chief executive, Airbnb, earnings call, 6 August 2026

“The customer service cost for booking is down about 16% year over year, in large part due to the AI agent.”

Elinor Mertz, chief financial officer, Airbnb, earnings call, 6 August 2026

Software development is where banks and payment networks give numbers. Bank of America (BAC $50.52 -0.08%) said on its January 2026 call that AI coding had cut the work of introducing a product or change, and Visa (V $404.16 -0.92%) gave the most detailed figures on its July call.

“We've taken 30% out of the coding part of introducing a new product to service or change that saves us about 2,000 people.”

Brian Moynihan, chief executive, Bank of America, earnings call, 14 January 2026

“And the results are meaningful for those teams that are using the agentic tool chain, with 80% more code commits and 80% plus improvement in requirement definition from 30 days to 5 days, which has translated to 65% plus faster feature development.”

Ryan McInerney, chief executive, Visa, earnings call, 28 July 2026

Revenue figures are the rarest. Medtronic (MDT $94.79 -0.65%) is one of the few to put AI into its growth rate, on its September 2026 call.

“Our coronary business was up low double digits, with CathWorks, our AI and advanced computational science platform for angio-based FFR, contributing nearly 300 basis points of organic growth.”

Thierry Pieton, chief financial officer, Medtronic, earnings call, 1 September 2026

The executives who won't

Some of the plainest statements on these calls were refusals to promise savings. On JPMorgan Chase (JPM $334.04 +0.74%)'s April 2026 call, an analyst asked whether the productivity boost from AI could meaningfully improve the bank's long-term efficiency ratio.

“I think it's a bad idea to think you're going to deploy AI and improve your efficiency ratio because in the competitive world, I'm going to do it, everyone else is going to do it, and the benefits will be passed on to the marketplace. It's not like you're entitled to have your ROE go to 50% and that'll stay there because you do it better than everybody else.”

Jamie Dimon, chief executive, JPMorgan Chase, earnings call, 14 April 2026

“But if you said, can you cut your cost profile by 5% tomorrow using all things AI, the answer is no.”

Strauss Zelnick, chairman, Take-Two Interactive, earnings call, 6 November 2025

Expedia, which has given figures for its AI customer service, said on its August 2026 call that its AI shopping features were not yet selling more trips.

“Those are not driving conversion right now, but what we're finding is that you get over 60% more information about traveler intent, and that allows us to deepen the relationships with a traveler.”

Ariane Gorin, chief executive, Expedia, earnings call, 5 August 2026

Analysts start asking for the number

In 2024 an analyst asked about AI savings, productivity or returns on about 1% of non-tech calls. Last quarter they did on 6.7%, and in the first 86 calls of October, on 15.1%.

One exchange on U.S. Bancorp (USB $59.70 +0.20%)'s October 2026 call shows the gap. The analyst noted that the bank had raised its operating leverage guidance by 100 basis points and had named AI as one reason.

“And that's a little surprising because it's hard for banks to actually quantify benefits and banks usually talk about AI benefits in more general terms and productivity and stuff like that. Were you talking in a more general sense, or do you actually have numbers around it?”

Mike Mayo, analyst, Wells Fargo, on U.S. Bancorp's earnings call, 15 October 2026

“So I think AI is, one, not the only reason to drive good productivity, but it is an important lever and has growing momentum and growing use cases for productivity within our world.”

Gunjan Kedia, chief executive, U.S. Bancorp, earnings call, 15 October 2026

The answer went on to the cost of AI models and the bank's confidence in “positive returns from our AI program on productivity”, without a figure.

Methodology

Source: speaker-by-speaker transcripts of earnings calls in the Equibles earnings call archive. Only calls classified as earnings calls are counted.

Sample: the companies in the S&P 500 at the latest index snapshot, and every earnings call they held with a transcript of at least 1,500 words from 1 January 2024 to 23 October 2026. Where a company has more than one transcript in a quarter, we keep the longest. Six calls are excluded because the same transcript is stored on calls in two different quarters.

Outside tech: we exclude companies in the technology sector, Alphabet and Meta (internet content and information), and Amazon, Tesla and AppLovin, whose sector is consumer cyclical or communication services but whose products are built around AI. That leaves 4,335 calls by 403 companies. Sectors are the company's classification on Equibles.

AI mention: a company speaker, meaning an executive, chair or investor relations officer, said “AI”, “A.I.”, “artificial intelligence”, “generative AI”, “GenAI”, “agentic”, “LLM” or “machine learning”.

Number on AI: a company speaker attached a specific figure to what AI delivered for the company or its customers: money saved, time or cost cut, a share of work automated, a productivity gain, or revenue from AI products, either achieved or as a stated target, with AI named as the driver or as one of the named drivers. Not counted: numbers of users, prompts, agents or tools; spending on AI; sales driven by AI demand, such as data center equipment or power; and statements of what AI could do. We searched every company speaker's sentences that mention AI for a number together with a benefit or revenue word in the same or the next sentence, and read each one. A second, wider pass over every AI sentence with a number added 36 calls. Each call counts once, whatever the number of figures on it.

Analyst questions: a sentence by an analyst that mentions AI together with savings, productivity, efficiency, returns, ROI, payback, margin or headcount. This is a keyword count, spot-checked by reading a sample.

Periods: the calendar quarter in which the call was held. Where a transcript carries only the end of the fiscal period instead of the call date, we count the call in the following quarter. October 2026 is reported separately because the reporting season is not over.

Limitations

Deciding whether a statement puts a number on AI is a judgment, and we made it by hand. A company that credits a cost program to AI and three other things counts the same as one that credits a result to AI alone; we report the split for last quarter.

The figures are what executives said on the call, not audited results, and a target is not a result. Some companies quantify AI in investor day presentations, press releases or annual reports rather than on earnings calls, and those are not counted.

The sample is today's S&P 500, so companies that left the index since 2024 are not in the early quarters. Sector samples are small; one or two calls move a sector's share by a few points. The speaker role is missing for some turns, and those turns are not counted as company speakers.

Conclusion

Outside technology, AI went from a topic on a third of S&P 500 earnings calls to a topic on most of them in under three years. Putting a number on it has not kept pace: last quarter executives gave a figure on about one in nine of the calls where they talked about AI, and the figures come mostly from a small group of companies in customer service, claims, software and company-wide cost programs. Analysts are asking for those numbers more often, and the most common answer is still that AI is one lever among several.

Every company's earnings call transcripts on Equibles come with speaker names and roles, and the free Equibles MCP server can search a company's recent calls, so an AI agent can check what a single company has said about AI and whether it gave a number. Our earlier study, Agentic AI Has Overtaken Generative AI on Earnings Calls, tracks which AI terms S&P 500 companies use.