I attended the 2026 Decorator’s Apparel summit by Printing United in Nashville in the end of June, and this is an excerpt of Davis Slagle’s presentation

The bottleneck is a person

Every shop has a ‘person.’ The one you send the hard design to, the rep who just knows what closes. That knowledge never leaves their head. Davis Slagle of Bee Graphics reframed it: it’s a knowledge-transfer problem, not a technology problem. The question that came out of it, how do you scale the thinking without scaling the person? Scaling expertise with AI.

Why they buy, not what

His team stopped cataloging what customers order and started documenting why. Four avatars, same hoodie, four reasons: the coach buying team pride, the business owner buying visibility, the booster president buying a fundraiser, the HR director buying culture. Coach Chris doesn’t want a hoodie. He wants the parents in the stands losing their minds on Friday night. The hoodie is the vehicle. They built those avatars, age, motivation, pain points, buying triggers, into custom GPTs, then wired years of sales and fundraising history together through Claude API calls and their own internal libraries. They didn’t teach AI to sell. Then They taught it how their best people think.

Moving design upstream

Then the part every decorator recognizes: “make something cool, black shirt, red print.” Proof, revision, revision, hair loss. Slagle moved design ahead of the designer. Using Wispr Flow to dictate into ChatGPT, a rep generates three on-brief concepts: edgy/athletic, feminine script, old-school, before the first call even ends. Not production art. A starting point. The designer still does the real work. They just start from something instead of nothing.

The real discovery was inward, Scaling expertise with AI

The surprise was understanding existing customers better. Pull the e-commerce, bulk orders, CRM, and store performance together, and Coach Chris stops being a hoodie order: he also needs senior gifts, a team store, a sponsor program. They knew what he bought. They didn’t know everything he needed.

What this really shows: the AI here isn’t doing the selling or the designing. It’s a distribution mechanism for expertise that’s used to tap into two or three irreplaceable people. The win isn’t the tool, it’s that the best thinking in the building finally scales past the person who holds it.

A note on agentic

Someone asked me in the Q&A whether the industry is moving from prompts to agents. Not yet. I build the behavior in code first. When I gave my art-generation tool an agent, it started forming opinions on its own images, generating one, going back to look at it, telling me what it thought. Not what I asked for. I treat it like building neural pathways: prove accuracy, run hard tests, earn trust, then open up the autonomy. Agentic is the goal. Handing it the keys today is not a safe space yet.

Tools used: Claude (API orchestration + internal libraries), custom GPTs (avatar storytelling), Wispr Flow (dictation), ChatGPT (proof-concept generation), Scribe (screen-capture clips), CRM (lead intake + segmentation).