The $300 AI Stack and the $3,000 Retainer

The Shopify services market is changing faster than most people in it want to admit.

For a lot of small brands, a $300/month AI stack is now delivering more value than a $3,000/month agency retainer. I've spent seven years in this ecosystem — agencies, freelance, apps, products — and that sentence would have been absurd three years ago. It isn't now.

The structure that's breaking

The old arrangement was tidy:

• Small brands hired freelancers
• Mid-size brands hired agencies
• Big brands built in-house teams

That worked because execution was scarce. Building a decent section, wiring up an integration, writing product copy at volume — each of those took skilled hours, and skilled hours had to be bought.

A lot of that scarcity has evaporated. Not all of it. But enough that the middle tier is under real pressure.

If your retainer was mostly throughput — a queue of small builds, copy updates, routine fixes — the value of that has genuinely fallen. A competent operator with good tools now clears a surprising amount of it in-house. Pretending otherwise is how agencies lose clients slowly and then all at once.

What didn't get commoditised

Here's the part I think gets missed in the doom-posting.

Demand for high-skill technical work went up. AI-integrated and headless Shopify work commands more now than it did, not less. The same forces flattening routine execution made the hard problems more valuable, because more people are attempting ambitious things and hitting walls.

What holds its value:

• Judgement about what to build. AI will happily help you build the wrong thing very efficiently.
• Architecture that survives contact with growth. Anyone can generate a section. Fewer can tell you why your theme will fall over at 4× traffic.
• The messy integration work. Systems that were never designed to talk to each other, edge cases, migrations with real data and real orders.
• Measurement. Knowing whether any of it worked.

That's a different job from the one a lot of agencies were selling. It's also a smaller-headcount, higher-skill job, which is uncomfortable for anyone who scaled on the old model.

So what should a brand actually do

If you're a founder trying to work out where to spend:

Under roughly $1M in revenue — buy tools, not retainers. Spend the money on inventory and traffic. Learn your own stack. You will be a much better client later for having done it.

$2M–$30M — this is where the honest answer gets harder. You have enough complexity that tooling alone stops being sufficient, but not enough to justify a full in-house team. What you need is usually not "an agency" in the throughput sense. It's a small number of people who can decide what matters and then build it.

Above that — you're hiring in-house and using outside help for specialist work.

The test I'd apply to anyone pitching you: can they build what they're recommending? If the answer is that strategy and execution live in different companies, you're paying for a translation layer, and that layer is where most growth plans quietly die.

The uncomfortable bit for people like me

This is not a comfortable argument for someone who sells services. It cuts against a chunk of the market I've worked in for seven years.

But I'd rather say it plainly than keep selling the old shape and watch clients work it out on their own. The version of this business that survives isn't the one that does more hours. It's the one that's worth more per hour — because the thinking is good and the thing actually ships.

This thinking shaped how we structure engagements — a paid audit sprint first, and we'll tell you when the answer is that you don't need us.