Executives at non-tech S&P 500 companies mentioned AI on 234 of 387 earnings calls in the third quarter of 2026. Only 25 calls attached a specific number to its benefits.
That gap sits at the center of an Equibles study published October 26, 2026, covering 4,335 calls from 403 companies between January 2024 and October 23, 2026. The share of calls mentioning AI rose from 32.1% in the first quarter of 2024 to 60.5% in the third quarter of 2026, nearly doubling.
The more consequential finding is what qualifies as a payoff figure. Equibles counts achieved benefits and targets, including company-wide savings programs where AI is one of several contributors. Its 6.5% headline therefore measures quantified disclosure, not the share of companies demonstrating a realized financial return.
For investors trying to separate AI industry enthusiasm from operating results, that distinction matters more than the frequency of the word itself.
AI Mentions Grew Faster Than Useful Disclosure
Equibles counted a call as mentioning AI when an executive used the term. It classified a call as quantified when an executive attached a specific figure to AI-related savings, productivity, automation, revenue or another covered benefit, whether achieved or targeted.
The study’s quarterly table shows how those two measures developed:
| Calls Held | Total Calls | Calls Mentioning AI | Calls Quantifying AI |
|---|---|---|---|
| January–March 2024 | 390 | 125, or 32.1% | 7, or 1.8% |
| January–March 2026 | 376 | 239, or 63.6% | 19, or 5.1% |
| April–June 2026 | 384 | 232, or 60.4% | 26, or 6.8% |
| July–September 2026 | 387 | 234, or 60.5% | 25, or 6.5% |
Quantification improved substantially from a small starting point. Among calls mentioning AI, the proportion including a number rose from 5.6% to 10.7% between the first and last full quarters shown.
Still, subtracting the 25 quantified calls from the 234 mentioning AI leaves 209 calls without a payoff figure in the third quarter of 2026. That is 89.3% of AI-mentioning calls.
These are call-level counts. A company can appear in several quarters, so 25 quantified calls should not be read as 25 companies proving an AI investment worked. The numbers track what executives disclosed during earnings calls, not every deployment or internal measurement.
What the Quantified Claims Actually Describe
Equibles grouped the 70 quantified calls from January through September 2026 into six categories. The largest was company-wide savings or productivity programs.
| Category | Quantified Calls | Share of the 70 Calls |
|---|---|---|
| Company-wide savings or productivity programs | 23 | 32.9% |
| Back-office work, including claims and documents | 17 | 24.3% |
| AI products and AI-assisted sales | 12 | 17.1% |
| Customer service | 8 | 11.4% |
| Software development | 7 | 10.0% |
| Product development | 3 | 4.3% |
Categories and counts come from Equibles; percentages are calculated from its 70-call total.
Two implications follow from that breakdown.
First, roughly one-third of quantified calls described broad company programs. Equibles says the most common figure was a savings target for a company-wide cost program with AI as one lever among several. Such a target can tell investors something useful about management’s plans, but it cannot isolate AI’s contribution.
Second, only 12 calls fell into the AI-products-and-assisted-sales category. The remaining 58 were classified elsewhere, predominantly around savings, productivity and operational work.
That does not make the other categories less valuable. Faster claims processing or lower customer-service costs can materially improve a business. It does mean the evidence is weighted toward efficiency claims, not a broad wave of separately identified AI revenue.
The 12 sales-related calls also should not be treated as 12 demonstrations of profitable AI products. The category includes AI-assisted sales, and the study’s overall definition allows targets. Neither the category label nor a revenue figure alone establishes incremental profit.
A Number Is a Starting Point, Not an ROI Calculation
Equibles’ broad definition is useful for detecting whether executives have moved beyond general statements. But investors need to distinguish three different disclosures:
- A measured operating result: time saved, work automated or a reduction in processing cost.
- A financial result: attributable revenue, lower expenses or a change in profit.
- A future target: an expected outcome that has not yet been delivered.
A time-saving result can be real without producing an equivalent cash saving. Employees may use recovered hours for additional work, quality checks or tasks that previously went unfinished. That can still be valuable, but multiplying hours saved by wage rates would require assumptions about how the business uses that capacity.
A broad savings program creates a separate attribution problem. If it combines AI, process redesign and other operational changes, its total target is not an AI-only return. Investors would need a breakdown before assigning that whole amount to AI.
Revenue requires another step. An AI-related product can generate sales while also incurring development, inference, infrastructure and support expenses. Revenue measures commercial activity; return on investment requires costs and a defined period too.
I’d read the 6.5% as a disclosure threshold, not a success rate. The study supports that interpretation because its definition deliberately mixes targets and achieved benefits, plus financial and operational measures. It does not provide a separate headline rate for realized, AI-attributable net returns.
The concentration of disclosure reinforces that caution. Across the full study period, 86 companies quantified AI at least once. Nine companies, led by Airbnb and UnitedHealth, accounted for 47 of the 158 quantified calls, according to Equibles. That is about 30% of quantified calls from a small recurring group.
Repeated disclosure can help investors follow progress. It should not be mistaken for equally broad disclosure across the sample.
Financial Services Leads, but Sector Comparisons Need Care
Financial services companies mentioned AI on 89.1% of their calls during January–September 2026 and quantified it on 11.4%, the highest rates reported in the study.
Energy and basic materials companies mentioned AI on 50% and 42% of calls, respectively, but neither sector produced a quantified disclosure under Equibles’ criteria during that period. Those are nine-month sector figures, not third-quarter-only comparisons.
The sector split shows why a single market-wide percentage needs context. Financial services’ high mention rate coexists with a much lower quantification rate. In energy and basic materials, AI appeared in executive discussion without a qualifying benefit figure.
That is a difference in disclosure behavior. It cannot tell us which sector has the best underlying projects, because earnings calls are a selective account of business activity.
The same issue affects the study’s October figures. Quantification reached 9.3% across the first 86 calls through October 23, but those calls were mostly banks and other financial companies. Equibles explicitly warns that this partial sample is not comparable with a full quarter. Treating it as an acceleration from 6.5% would overlook the changing mix of companies.
Frequently Asked Questions
3 questions
1What Did Equibles Analyze?
The study covered 4,335 earnings calls from 403 non-tech S&P 500 companies, held between January 2024 and October 23, 2026.
