The MLSE CEO brought AI-generated trade returns to the Leafs war room. The story isn’t that he used the tool. It’s what it means that he needed to.
On March 6, the day of the NHL trade deadline, Keith Pelley sat in the Toronto Maple Leafs war room. He had told reporters at a March 31 press conference that he was no more than a silent observer. The Athletic’s Chris Johnston reported something different. Pelley peppered scouts with questions, demanded more assets in trade conversations, and tossed out opinions and ideas.
He also arrived with notes on possible trade returns that Leafs staff members believed were generated by large language models — tools that had become increasingly influential at the top of MLSE but had not been part of the usual process with its NHL team.
When Pelley addressed it publicly, at the press conference introducing John Chayka as the team’s new general manager, he called it preposterous. There have been countless erroneous reports and inaccurate information, he said. He did not want to comment on them, including the one about using AI at the trade deadline. Preposterous.
Who Sent Him There
Johnston’s reporting identified the person behind the directive: Humza Teherany, MLSE’s chief strategy and innovation officer, and one of Pelley’s most trusted advisors. Teherany was not in the room. His influence, Johnston wrote, loomed large. The AI tools Pelley brought were not something he developed independently or snuck in. They were part of a broader push by MLSE’s leadership to make large language models part of how the top of the company thinks and operates.
Teherany’s title is worth pausing on. Chief strategy and innovation officer is a role that exists to translate corporate priorities into institutional practice. At MLSE, the corporate priority since Rogers announced its buyout intent is asset value. Staffieri said so explicitly: the company held significant sports assets and was getting no credit for them in its share price. The plan was to consolidate and surface that value. That is not a hockey goal. It is a balance sheet goal. Teherany’s job, and by extension Pelley’s presence in the war room, is the balance sheet expressing itself through hockey operations.
What the Denial Reveals
Calling the report preposterous is not the same as denying the facts. The detail — notes containing possible trade returns that staff believed were LLM-generated — was confirmed across multiple independent outlets quoting Johnston’s piece directly. Pelley did not sue. He did not produce a corrective account of what the notes actually contained. He called the characterisation erroneous and moved on to discussing Chayka.
The preposterous is about optics, not accuracy. What Pelley found intolerable was not the use of the tool but the image the story produces: a non-hockey executive, in a hockey operations room, on the most consequential single day of the hockey calendar, working from machine-generated output. That image is damaging not because it is false but because it is legible. It makes visible something that is supposed to stay in the background.
The image is: a CEO who does not have hockey judgment, using a tool that simulates having it, in a room full of people who do have it, on a day when it matters most. That is not a story about AI in sports. That is a story about who is actually running the institution and what they are running it for.
Deskilling Is Not About the Tool
The argument that AI is deskilling sports management has been building for several years and is usually framed as a technical concern: algorithms replacing scouts, models replacing intuition, data replacing judgment. That framing puts the problem in the technology. The Pelley scene puts it somewhere more useful — in the institutional purpose the technology is serving.
Pelley is not using LLMs because they are better at evaluating trade returns than Leafs scouts. He is using them because he does not have the hockey knowledge to generate his own read, and generating his own read is not actually his job. His job is to represent the parent company’s interests in hockey operations decisions that affect asset value. A tool that produces a plausible-looking set of trade parameters lets him perform hockey judgment without developing it. That is useful to a company that wants a non-hockey CEO running a hockey team.
What gets displaced in that arrangement is not just expertise but the conditions under which expertise matters. If the person with final authority in the room is working from LLM output, the scouts and analysts whose years of judgment produced the actual read are not informing the decision — they are providing the raw material the algorithm processed to produce the summary that arrived in a non-hockey executive’s notebook. The knowledge still exists in the room. It has simply been routed around.
What the Company Is Actually For
The deadline went badly by any hockey measure. The Leafs had two low-cost rentals to move in Laughton and McMann. With thirty minutes left, Treliving emailed all thirty-one other GMs to inform them Laughton was available. Rival executives found the Leafs difficult to deal with throughout — calls felt like tire-kicking, and promising conversations collapsed in the final stages. The team received five draft picks for its deadline activity. Treliving was fired shortly after.
The hockey result and the corporate result were not the same event. During those eleven days that spring, MLSE’s two marquee clubs made their largest moves in years while Rogers completed a CAD $4.35 billion purchase at a thirty-nine percent premium. The deadline was one of those moves. The picks were assets. A non-hockey CEO in the room with LLM-generated returns is not a scandal about the technology. It is what asset management looks like when the asset is a hockey team and the manager does not need to know hockey to manage it.
Pelley fired the general manager who was in the room that day and replaced him with John Chayka. He introduced Mats Sundin as a senior executive advisor. The press conference was well-attended, the names were credible, and the organisation appeared to be doing something. That appearance is also an asset. The preposterous is how you maintain it when the underlying scene has already been reported.
Sources
- The Leafs Nation — summary of The Athletic / Chris Johnston (April 15, 2026); Pelley’s presence, the LLM trade returns, the Teherany directive, rival executives’ frustration, the Laughton email with 30 minutes to go
- Maple Leafs Hot Stove — independent analysis; Pelley’s original March 31 claim to be a silent observer; confirmation that his active presence was already contested before the piece ran
- The Leafs Nation — Pelley’s denial at the Chayka introductory press conference; the “preposterous” quote (May 2026)
- Yahoo Sports — Pelley brought AI trade pitches to deadline meeting; Johnston quote reproduced; Pelley’s subsequent denial noted
- Sportico — Rogers CAD $4.35 billion Kilmer purchase (July 6, 2026); Staffieri on surfacing asset value for shareholders