Skip to main content
Paul Welty, PhD AI, WORK, AND STAYING HUMAN

The historic opportunity

Substack article draft. Ad register — punchy, provocative, first-person. Run through /humanize before publishing.


The historic opportunity

Claude Design shipped on April 17. I changed my whole business on April 21.

Here’s what happened in four days. Anthropic released Claude Design — “designs, prototypes, slides, one-pagers, and more” — bundled into the Claude subscription a lot of companies already pay for. Canva previewed AI v2 on the CEO’s podcast the same week. GitHub rolled Copilot Workspaces deeper into their product. Google shipped Gemini 3 into Workspace. Microsoft kept pushing ChatGPT into Excel and Word.

The pattern is clear now. Every major AI vendor is bundling execution into the tool. Not assistance. Not copilots. Actual production work — the kind of work agencies have been selling by the asset for thirty years.

I watched that happen and I rewrote the Synaxis offering the next weekend.

What most consultants are going to miss

The reflex is to call this a squeeze on agencies, declare that some firms will die, and write a think piece with a lot of sober nodding. That’s the view from inside the agency. It’s also the least interesting view.

Try the other side of the table. Sit where the client sits. You run marketing inside a mid-market firm. You have a team of three, maybe four. You’ve been hiring an agency for two years because you needed real design, real positioning, real content, and your team couldn’t produce at that level on their own. That was true in 2023. It was true in 2024. It was still true last year.

It isn’t true now.

Your team, plus Claude Design, plus Canva AI, plus ChatGPT embedded in the tools they already use — that stack can produce at agency standard. Not “close to.” Not “good enough for a startup.” Actual standard. First time in history that’s been possible. The cost curve flipped in about eighteen months and most people haven’t caught up to what that means for how the work actually gets made.

The loss for agencies is the gain for every internal marketing team that’s been quietly working above their station for a decade.

The thing nobody tells you

The skills don’t install themselves.

You can hand a team Claude Design and Canva AI and ChatGPT and they will produce exactly what they produced before — only faster and with more polish on the surface. Garbage in, glossier garbage out. The tools don’t fix the taste. They don’t fix the judgment. They don’t fix the structural stuff — positioning, story, brand architecture, the quiet decisions that separate a strong one-pager from a forgettable one.

So there’s a sequence that matters, and most people are getting it wrong.

AI skills come first. You can’t teach marketing craft to people who haven’t absorbed the tools yet, because the tools change which questions are worth asking. A team that fully understands what Claude Design will and won’t do is a team that asks different strategic questions than a team still treating it as a toy. Get the tool fluency in first. Then teach the craft on top of it.

That’s the educational structure that actually works. I’ve been testing it.

Twenty-five years

I’ve been doing this work for twenty-five years. Not consulting about it — doing it. Building marketing systems inside Fortune 500 companies. Running innovation centers. Shipping products. Running forty-three AI experiments in an 18-month stretch at Emory before most firms had a ChatGPT policy. The methodology I’ll hand your team isn’t a framework I cooked up last quarter. It’s the one I used to ship real work for real organizations under real deadlines.

What I never had before was this moment. A moment where the tools finally caught up to the ambition most in-house teams already have. Where I can walk into a firm, teach the team directly, and leave behind not a dependency but a capability. Where the output of the engagement isn’t a deliverable that walks out the door with me — it’s a team that now performs at a level they couldn’t perform at six months ago.

That wasn’t possible in 2022. It is now.

One door

Synaxis used to offer marketing services and AI implementation. Two lines. Four tiers. The usual agency spread.

I collapsed it into one offering on Monday.

Teach the team. Take the opportunity. That’s it. That’s the whole commercial surface.

I’ll come in, assess where your team actually is with the tools, install the AI skills first and the marketing craft second, and leave behind a team that produces at a standard that used to require outside help. You stop paying an agency to make your assets. You stop paying a SaaS tool to make them generic. Your team owns the work.

The offer

We’re at a moment in history when AI has made something new possible: your internal team should be doing their own marketing and design work. I’ve done this work for twenty-five years. I can help your team learn the AI skills first, and then the marketing and design craft they’ll need to take real advantage of this historic opportunity.

If that sounds like the shape of your firm, reply or email me.

One door. That’s the whole pitch.

The agent-shaped org chart

Every real org has the same topology: principal, role-holder, specialists. Staff AI maps onto it, node for node, and the cost collapse shows up in the deliverables that were always just human-handoff overhead.

AI as staff, not software

Two frames for what AI is doing to work. The tool frame makes tools smarter. The staff frame makes roles unnecessary. Those aren't the same product, the same company, or the same industry.

Knowledge work was never work

Knowledge work was always coordination between humans who couldn't share state directly. The artifacts were never the work. They were the overhead — and AI just made the overhead optional.

The work of being available now

A book on AI, judgment, and staying human at work.

The practice of work in progress

Practical essays on how work actually gets done.

Shopping is the last mile

Every meal planning app treats cooking as the hard problem and shopping as a logistics detail. They have it backwards. Cooking is mostly solved. Shopping is the last mile.

Watch what they buy, not what they say

Forms ask people to declare preferences. Receipts record what they did. The gap between the two is where revealed preference lives, and it's wider than most product teams admit.

What the API decides not to show you

Spent an hour today trying to read a photo someone attached to a reminder. The bytes are right there on disk. Apple won't let me see them. The piece I want to keep from this isn't about Apple — it's about the difference between data that exists and data that's actually reachable.