Five Lessons From RetailClub AI Festival for Distributors Getting Ready for AI Buyers

Three days in Huntington Beach, and one line nearly every retail leader repeated in some form: start with the problem, not the technology. Walmart’s Tracy Poulliot, Sam’s Club’s Greg Pulsifer, and Amazon’s Amanda Doerr all made the point. At an event built entirely around AI, the most consistent advice on stage was to stop leading with AI.

The 2026 RetailClub AI Festival was a retail conversation, but from our perspective most of the insights and advice also applies to B2B distributors and manufacturers. These five themes kept coming up, and each one has a B2B translation.

1. Your product data is the storefront agents actually see

No theme came up more often. Whatever the surface, ChatGPT or Meta’s new Muse agent, an AI agent can only recommend what it can read, and most catalogs were written for people scanning a page.

Anshuman Taneja, Chief Digital and Technology Officer at GNC, described spending six months building a single golden record for its product data before putting a customer-facing agent in front of shoppers, because scattered ingredient and dosage data produces wrong answers. At David’s Bridal, ChatGPT once answered “unknown” when asked if the company had stores, CTO Scott Saeger said, because its store pages were tagged incorrectly in the sitemap. For fashion, Josh Krepon of Steve Madden noted that a shirt record reading “cotton, made in Peru” gives an agent nothing to reason with. Estée Lauder’s Melina Flabiano said the team now designs every page for two customers: a human and an agent.

Data from outside the event agrees. In Q1 2026, AI-referred traffic to US retail sites grew 393% year over year, yet around 34% of product pages could not be properly accessed by AI (Adobe Analytics, April 2026, via TechCrunch). In her post-event recap, Nikki Baird of Aptos named the underlying issue: AI exposes existing data gaps faster, and may raise the cost of leaving them unfixed.

The B2B read: distributor catalogs have the same problem at a larger scale. Specs, compatibility, units of measure, substitutes, and contract-specific availability usually live in PDFs, ERP fields, and reps’ heads. When a buyer asks an AI engine for a part that fits a specific application, the distributor with structured, complete attributes gets the recommendation. We have written about how AI changes B2B product discovery, and this event reinforced the order of operations: product data work comes before agent work.

2. Agentic commerce is a range of delegation, and buyers pick their spot

The fully autonomous shopping bot got less airtime than expected. Rajiv Mehta, Amazon’s VP of Search and Conversational AI Shopping, described a spectrum: shoppers happily hand off chores (reordering the same socks every three months) but stay hands-on for purchases they enjoy or that carry risk. Poulliot and Target’s Sarah Travis both described shoppers moving away from item-by-item shopping toward shopping by mission, such as dinner tonight or a home repair. Travis said AI works best for Target when it can build a full basket for that mission.

The B2B read: B2B starts with an advantage at the easy end of that range. Replenishment, standing orders, and repeat buys against contract pricing are the low-stakes, high-certainty purchases buyers will delegate first. Missions translate too. A contractor is not buying one fitting; they are kitting a job, and the distributor that can assemble the whole job from a description, at the right account pricing, is ready for how B2B buyers will delegate.

3. The work that matters sits underneath the agent

Target had the most striking architecture story of the week. Prat Vemana, Target’s Chief Information and Product Officer, said roughly two-thirds of the Target app is new code written in the last 18 months, rebuilt on a composable architecture so prediction models can assemble each shopper’s page. Vemana was blunt that putting an MCP server in front of an existing API and calling it an agent misses the point. Systems built for humans to read were never built for agents to act on.

The B2B read: the real question is whether your pricing engine, entitlements, inventory, and order management can answer an agent accurately and in real time. We made that argument in Your AI Pilots Aren’t Failing. Your Commerce Layer Is., and our agentic commerce readiness work begins with that layer.

4. The measurable wins are in operations

Baird also observed that most AI change so far sits on digital and ecommerce surfaces rather than deep in operations. Yet the results people could put a number on came from the back office. Ekta Chopra, Chief Technology and AI Officer at e.l.f. Beauty, said a payroll consolidation process went from four weeks to four hours. At Sam’s Club, a merchandising task that took four associates half a week now takes one person half a day, Pulsifer said. Tuckernuck founder September Votta credited an in-house AI analytics tool, paired with smarter marketing signals, with lifting gross margin three points in 60 days.

McKinsey partner Jacob Ader’s maturity research, presented on day one, explained why this matters. Most employees say AI makes them more productive, yet only a small share of companies can tie a meaningful EBIT gain to it. The companies that can are redesigning whole workflows instead of adding point tools.

The B2B read: for distributors, the equivalent workflows are quoting, order entry from emails and PDFs, pricing analysis, and demand forecasting across branches. They are unglamorous and measurable, and they sit closer to margin than any chatbot. Each depends on the same foundation as the first theme: product, pricing, and order data that is clean and connected. That foundation is where our data and AI readiness work starts.

5. Adoption is a people problem with a governance answer

Operator panels kept returning to change management. Companies with results described leaders using the tools visibly, broad employee access through a governed front door, and a clear rule for where a human reviews the output. ThredUp co-founder and CEO James Reinhart put it plainly: building agents is easy, and making them create business value is the hard part. Two cautions came up: token costs climb faster than teams expect, and depending on one model is risky.

The B2B read: B2B has an extra constraint. Sales reps, customer service, and branch teams hold the account knowledge, so adoption has to include them. Part 4 of AI Best Practices for Commerce, our free online reference, covers adoption and implementation in detail.

What we are taking home

The retailers furthest ahead did not start with the most impressive agent. They started with clean product data, a commerce layer an agent can transact against, and one operational workflow where the result could be measured. Distributors and manufacturers can do the same work now, before AI-assisted buying becomes a meaningful share of B2B orders.

To see where you stand, run the free McFadyen AI Commerce Readiness Audit. It shows how AI engines and agents read your site and catalog today, and where the gaps are.