A Goldman Sachs (NYSE:GS) M&A banker just put a number on something most AI commentary leaves as a feeling. Speaking on Bloomberg’s Odd Lots podcast in an episode dated September 19, 2026, Gene Sykes said only 2% of companies have reported a positive impact on earnings per share from AI in their earnings reports. In the same conversation, Sykes acknowledged that roughly 20% to 30% of companies have implemented AI strategies.
That distance between adoption and impact is the story worth sitting with. Strategy decks are one thing. A line item that moves EPS is another.
How the Number Came Up
Co-host Joe Weisenthal floated a deliberately provocative theory that enterprise AI adoption is “literally zero,” arguing that handing employees a chatbot subscription amounts to glorified search rather than meaningful integration. That framing was the setup for Sykes to push back with the figures above. The pushback reframed the hype cycle. A meaningful minority of companies are doing something with AI. A tiny fraction have anything to show a shareholder for it yet.
For investors listening for a way to size the gap between corporate press releases and quarterly filings, this is the cleanest practitioner data point that has come out of a major bank all year. It is worth reading the exact language in the episode itself.
What Sykes Says Real Integration Looks Like
Asked what genuine integration looks like, Sykes pointed at his own employer. He described Goldman Sachs as deeply embedding AI into workflows and said its youngest employees are “native AI people” implementing things the firm was not thinking about a year ago. That is his characterization of his own firm, so read it as such rather than as an outside audit.
The more interesting exchange came when co-host Tracy Alloway pressed him on what AI actually does inside M&A, a business built on relationships. Sykes framed the technology’s value in efficiency terms. He said it makes advisors faster at getting to conclusions and “more imaginative,” giving them access to more information and better information, improving judgment and confidence, according to Goldman Sachs. He expected AI to preserve the relationship layer of dealmaking while raising the floor for advice quality across the industry.
A higher floor rather than a higher ceiling. That is a sharper claim than most AI-in-finance commentary offers, because it implies the tools will compress the quality gap between top advisors and everyone else before they extend the frontier.
Where AI Is Already Reshaping Deals
The contrast at the heart of Sykes’s segment is that AI is visibly reshaping who is buying companies while barely registering in reported earnings. He described a record-breaking M&A market and noted that private equity’s share of deals has dropped from 40% to 30% as AI-driven strategic demand surges across every industry, including semiconductors, according to Goldman Sachs. Strategic acquirers with AI roadmaps are outbidding financial sponsors for assets that plug into those roadmaps. The deal table shows the shift long before the income statement does.
That framing lines up with reporting elsewhere that corporate America is moving too slowly on AI, held back by a mix of caution and poor recruitment, per a recent Fortune piece. The friction is real, and it shows up in the gap Sykes is measuring. For investors trying to figure out where AI dollars are actually landing today, the picks-and-shovels layer, power, cooling, networking, is the part already showing up in orders (we profiled seven of those suppliers in a free report you can grab here).
What to Watch Next
The takeaway for individual investors is a measurement to weigh. A meaningful share of public companies say they have an AI strategy. A very small share can point to AI on the earnings line. Both things are currently true, and the useful exercise going into the next reporting cycle is asking which management teams are willing to attribute a specific dollar of margin, revenue, or productivity to AI, and which are still describing plans. Sykes has given the market a benchmark to hold those disclosures against.