AI and Productivity: What Firms Are Saying on Earnings Calls

-->

Search

AI and Productivity: What Firms Are Saying on Earnings Calls

July 31, 2026

By

Serdar Ozkan

Aakash Kalyani

Nicholas Sullivan

SHARE THIS PAGE:

Link Copied

KEY TAKEAWAYS

An analysis of 490,000 earnings call transcripts from publicly traded U.S. firms suggests AI has reshaped how executives talk about productivity, even if its effects are not yet clearly visible in aggregate productivity data.

Discussion about productivity has risen in recent years, but more striking is the sharp increase in the share of productivity-related discussion that centers around AI.

The majority of talk about productivity is forward-looking, especially when AI is involved. Approximately 95% of the productivity-related sentences in earnings call transcripts that mention AI refer to future gains.

Executives predominantly describe AI-related productivity effects in positive terms, suggesting a strong sense of optimism.

Since the release of ChatGPT in late 2022, artificial intelligence (AI) has been widely expected to raise productivity. But aggregate productivity data have so far offered a more muted signal: Utilization-adjusted total factor productivity grew only 0.07% over the four quarters ending with the first quarter of 2026.This reading, from the San Francisco Fed, reflects the June 4, 2026, update, which was current as of July 20, 2026. In this blog post, we use quarterly earnings calls to measure how business executives discuss productivity, AI and the timing of expected productivity improvements.

AI Productivity in Business Executives’ Own Words

Our analysis draws on about 490,000 earnings call transcripts from S&P Global covering 5,198 publicly traded U.S. firms from 2000 through 2025. Earnings calls are useful for this purpose because they occur regularly, cover a broad set of firms, and provide a comparable window into how managers discuss performance, costs, investment and expectations over time.

Finding and interpreting corporate productivity discussions inside millions of sentences of text require a two-step approach:

First, we scanned every earnings call transcript and extracted each sentence containing words related to productivity—such as “productive” and “efficiency”—along with the sentences immediately before and after for context. This step is intentionally broad; it will catch some irrelevant uses, such as a reference to a “productive meeting.”

Second, we passed each matched sentence and its context to an open-source large language model (LLM), Qwen3-3B, which extracted several key pieces of information. The LLM determined whether the discussion also involved AI; whether the firm described productivity as rising, falling or unchanged; and whether the speaker was describing something that had already happened, was happening then or was expected in the future. Sentences that turned out not to be about firm productivity were flagged and dropped.

The result was a sentence-level dataset of roughly 910,955 tagged productivity sentences, each linked to a firm and quarter. This let us measure not only how often firms discussed productivity, but also whether AI-related productivity discussions differed from other productivity discussions in timing, tone and substance.

The sections that follow discuss the results of our textual analysis.

The Share of Productivity Discussions Devoted to AI Has Risen Sharply

Before we turn to AI, it is worth noting that discussion in earnings calls about productivity overall has risen during the sample period. The first figure below shows the share of sentences per transcript flagged as productivity-related, smoothed using a four-quarter trailing moving average. The series dipped sharply in 2020, especially in the early months of the COVID-19 pandemic, when earnings calls were dominated by immediate concerns about demand, supply chains and labor-market disruption. As firms moved out of crisis-management mode, productivity discussions rebounded and then continued to rise. By 2025, productivity-related sentences accounted for a larger share of earnings call transcripts than they did before the pandemic.

The composition of productivity discussions has changed even more sharply. (See the next figure.) The share of productivity-related sentences that also mention AI was near zero before ChatGPT’s debut. It rose quickly in 2023, plateaued in 2024, and then accelerated again in 2025, reaching roughly 15% of productivity sentences by the end of that year. In other words, firms are talking somewhat more about productivity overall, but more strikingly, productivity discussions are increasingly being framed through the lens of AI.

Additionally, the way firms talk about AI has evolved. To illustrate this change, we calculated the frequency of certain bigrams (two-word phrases) in our collection of identified sentences. (See the following figure.) In 2023,...