Executives are increasingly talking up AI productivity on earnings calls

You’ve heard of name-dropping—but have you heard of productivity-dropping?

According to the Federal Reserve Bank of St. Louis’s analysis of nearly 500,000 earnings call transcripts between 2000 and 2025, C-suite leaders are spending more time on corporate earnings calls discussing productivity: The share of AI-related productivity sentences was 15.6% in Q4 of 2025, up from 1.5% in the same period in 2022, the year ChatGPT launched.

The St. Louis Fed was able to determine this by scanning transcripts for mentions of keywords related to productivity (e.g., “efficiency”) and using Qwen3-3B, a large language model, to determine details like whether those keywords were used to discuss AI.

More to come. When talking about AI productivity gains, executives often focused on the future delivery of benefits. Roughly 95% of AI-related productivity sentences were forward-looking, according to the analysis, compared to 74% of non AI-related productivity sentences that discussed future gains.

“Firms may be investing, experimenting and reorganizing around AI today, while the measurable productivity effects remain mostly ahead,” the report wrote. More than 9 in 10 (95%) AI-related sentences indicated productivity was going up.

Agree to disagree. While executives on corporate earnings calls have been bullish about AI productivity gains, sentiments from IT leaders tell a different story. A March report from technology service provider Logicalis found that while 94% of global CIOs said the “appetite for AI adoption” within their organization has increased in the last year, 51% believe AI implementation is going too fast. Another 55% said they weren’t confident that their organization had a clear AI roadmap for the next two to three years.

Real gains? Some studies have also called into question the true productivity from AI tools. IT Brew previously reported findings from a 2025 METR study that found AI tools actually slowed down experienced developers’ workflows. A separate study suggested that AI coding tools may help developers produce more lines of code, but don’t necessarily equate to more shipped software.