PPAI just published research on AI in branded merchandise, and the headline number is the kind you read twice: adoption is almost total. Ninety-eight percent of suppliers and nearly every distributor are already using or testing AI. In an industry that isn't always first to new technology, that's remarkable — and it means the interesting question has already moved on. It's no longer whether our industry uses AI. It's how deep it goes.
Read past the adoption numbers and the same study answers that, almost in passing. PPAI economist Alok Bhat described what's coming next as "a shift from individual productivity to business execution." That's the whole thing in six words. Because right now, nearly all of the AI in our industry is individual productivity — and there is a ceiling on how far that takes you.
The tab problem
Look at what firms actually use AI for in the study: marketing copy, sales emails, product descriptions, mockups, reporting. All genuinely useful. All of it also happens in a separate tab. You open ChatGPT or Claude, describe what you need, get a draft — and then, this is the part that matters, you copy it back into the system where the work actually lives. Your catalog. Your order. Your store. Your client's presentation.
That copy-paste step is the tell. It's why the number-one barrier in PPAI's own data isn't cost or skepticism — it's system integration, cited by more than half of distributors. The AI can't see your catalog, can't touch your orders, can't change your store. It can only advise from the outside. And the tools doing the advising — ChatGPT, Claude, Copilot — are general-purpose assistants, not the platforms the industry actually runs on.
This is what I'd call bolt-on AI: intelligence layered on top of systems that were never built to be operated by it. Bolt-on AI is real and valuable, and it's also where most of the industry is about to feel a ceiling. You can make every person a little faster at drafting. You cannot make the business run differently — because the AI never does the work, it only prepares it for a human to do.
Bolt-on vs. AI-native
Getting past that ceiling isn't a matter of adding more AI features. It's a matter of architecture.
The alternative to bolt-on is what people are starting to call AI-native (or agent-native): systems where the core actions aren't just clickable by a person — they're operable by an AI. In our world those actions are specific and concrete: finding the right product, embellishing it, configuring an order, spinning up a branded store, generating a custom report. In a bolt-on world, AI can describe those things. In an AI-native world, AI can do them — because the system exposes its real operations, not just its screens.
The difference sounds subtle. It isn't:
Bolt-on AI suggests, and a human executes. AI-native AI executes, and a human decides.
One removes minutes. The other removes the manual layer.
There's a second half to this that I'll come back to in a later piece, but it's worth planting now. The interesting version of AI-native isn't a smarter chatbot buried inside one more product. It's open. The same way REST APIs once let any application talk to any other, a new open standard is emerging that lets AI tools talk to software directly — so the AI you already use could reach the systems you already run. PPAI's data shows the industry has effectively standardized on a handful of AI assistants. The real question is whether your platform meets you inside them, or asks you to leave them.
When intent becomes the interface
Follow that far enough and the interface itself starts to change. For decades, using software has meant learning where the buttons are. AI-native software makes intent an interface too. "Build a store for this client with these ten products and their logo" becomes something you can say, not a sequence of screens you have to navigate. Over time, for the people who want it, AI becomes a new front door to the platform.
I want to be careful with that claim, because the hype around it is exhausting and mostly wrong. AI is not going to replace the interface, and it is certainly not going to replace people. The screens stay. The point is optionality — a faster path alongside the familiar one, for the moments and the users where it helps. Some people will live in it. Some will never touch it. Both are fine.
What AI can't do — and shouldn't
The same PPAI study has the best rebuttal to the hype, from Brand Fuel's Danny Rosin: "AI cannot replace trust." He's exactly right, and it's the reason AI-native matters more than it first appears. Trust is built in the human parts of this business — understanding what a client actually needs, solving the problem no brief captured, being the person who picks up the phone. Those are not the parts anyone should want to automate.
The mechanical parts are. The artwork handling, the order entry, the re-keying between systems — we've written before about the hidden cost per order those mechanics create, and why more stores shouldn't mean more work. AI-native isn't about replacing the relationship. It's about giving people back the hours the software currently eats, so more of their day goes to the work only they can do.
Where this goes
So here's where I think this is heading. The industry has cleared the first bar — everyone is using AI. The next bar is the one PPAI's own economist named: moving from individual productivity to business execution. And you don't get there by adding another AI feature to a system that was designed before any of this existed. You get there by rebuilding the system so that AI is native to it — so the platform's real work is available to the AI, openly, from wherever you already work.
That's a bigger project than a feature release. It's a different starting assumption about what software for this industry should be.
It's the assumption we're building on. More on the how — including that open standard I mentioned — in the pieces to come.