The Watson Weekly is sponsored by Avalara — the agentic AI platform automating global tax and compliance for leading eCommerce brands with simple plans, transparent pricing, and complete coverage across Shopify, Stripe, WooCommerce, BigCommerce, and more.

Andy Hoar spent about a year hunting for a thesis he could defend before he wrote Bot to Bot. What he landed on is that B2B has been putting its money into the wrong part of the journey. I've been chewing on the conversation since we recorded it.

The first thing is that the unglamorous middle is where the money is. Every budget I see goes to the storefront and the last mile, because those are the parts a customer can point at. Andy's argument is that B2B buyers have no interest in looking at your site. They're hunters, in his framing. They want in and out. So the work worth automating sits between the click and the dock, where pricing gets calculated, inventory gets allocated and orders get tracked, and where nobody is watching. His example is a distributor pulling a National Weather Service feed and rerouting orders out of a warehouse in a tornado's path, shipping the same product from a neighboring state instead. The customer never finds out any of that happened. What I like about this is that it gives you something to check on Monday morning. If nobody at your company can walk you through how an allocation decision actually gets made today, you're not ready to hand that decision to an agent.

The second is that "we have a data problem" is true and it's also an alibi. Andy asks rooms of B2B people whether they have a data problem and says every hand goes up, every time, with nobody arguing. I pushed on that during the interview, because a condition that universal starts doing a job other than describing reality. It becomes the reason to wait. His answer was better than my question. He says triage the data instead of trying to fix all of it, because some of it is already good enough for what you're actually trying to do. If your product descriptions are sitting around an 8.5 out of 10 across the catalog, another attribute or another photo isn't the thing standing between you and revenue. Wall off the parts that are genuinely broken, and come back when you know what you need them for. He also said something I'll be repeating. The default decision in most companies is to do nothing, and it's the only decision that never has to be defended, because it doesn't challenge anything anybody already did. Almost nobody runs the math on what the waiting costs.

The third is that the moat everybody assumes is permanent has an expiry date on it. Distributors have been selling expertise for a very long time. The person who knows which fitting is rated for what pressure, which part fits the older unit, what you forgot to order. Andy's question is what that expertise is worth on the day an AI advisor answers about as well as somebody with twenty years behind the counter. It doesn't have to be perfect to do damage. It has to be good enough that the customer stops picking up the phone. After that a distributor is competing with Amazon on price and delivery speed, which isn't a fight I'd want.

I'd add one thing to Andy's advice about who to trust. He says trust the person asking the best questions rather than the one with the best answers, and I agree with that. I'd also say go read your customer verbatims, the raw unedited comments nobody has tidied up, and then go find the accounts that buy one item from you and nothing else, and ask them why not. Your advisory board won't tell you. Your sales team staffed it with your happiest customers, and there's a quarterly steak dinner riding on it.