I’ve sat through plenty of AI sessions at retail events where the speaker is selling something or describing a best case that their own company hasn’t reached. For Retail Club this year I wanted people to argue a position in front of a room of operators and then find out how the room voted. We ran the first Watson Live! The Retail AI Debates at RetailClub Presented by KIBO Commerce with Ekyam and Power Digital as supporting sponsors.

I warned the audience up front that some of the positions would be a bit of a straw man. Each motion opened with three minutes for the affirmative, followed by shorter replies from both debaters and a question or two from the floor. People voted with the red and blue plates on their tables.

Kelly Goetsch argued against changing workflows around AI before the data is ready, which is something he believes, and he brought the most specific examples of the day. One was his account of Target at one point running seven separate systems of record for inventory. Gireesh Sahukar argued the other side, saying that bad data comes from a flawed business model and improves as the business does. He also asked the retailers and brands in the room to raise their hands, and of about ten, only one said their data was good. Gireesh won the vote easily.

Kevin Reese argued that companies won’t regret AI layoffs because the P&L will force restructuring and the mistakes will be in how it gets done. Casey Golden said retail has never written down how most of its jobs work, so the people who leave take that knowledge with them and get hired back later at a higher price. The room sided with Kevin. During audience questions, someone asked what a new graduate or a person five years from retirement should do about their job, and neither debater could offer much reassurance.

For the third motion I stepped in myself against Danny Smith from Stripe, who worked on Rufus at Amazon before that. I argued retailers should put their own agents ahead of optimizing for ChatGPT and Gemini, because the agent needs inventory and delivery data that I wouldn’t hand to OpenAI. Danny asked who had researched a purchase in a chatbot, and most of the room had. He counted four people who had used Rufus. I won the vote, though Danny had said beforehand that the host would.

Anatoli Iakimets, director of product marketing at KIBO Commerce, closed with a short fireside conversation. He said he agreed with both sides of the first debate, that retailers need to start on AI now and that most of them shouldn’t try to build advanced technology themselves.

Thank you to the debaters and to everyone who stayed for the votes.

I’m still thinking about the fact that a room which admitted its data isn’t ready voted for starting AI work now, and I haven’t decided what that says about how retailers plan to close the gap.