Give away the advice, charge for the build:a footnote on free AI demos

I've built three prospects something for free with AI, and my scorecard on those is 0 for 3. So here is my new rule, written down where I can't quietly break it: my fastest work is for paying clients, and a prospect who wants a demo gets a diagnosis instead.

Note to self: give away advice, charge for the build

In August I posted that you should pick your people and outgive them. I still believe it. This is a footnote, because I've been outgiving in the most expensive way I know: free builds.

AI makes the build very hard to resist. A prospect describes an idea on Tuesday, and by Thursday I've built it. Every time, I tell myself this is the demo that closes the deal.

It's an easy story to believe, because the build is the fun part. Writing a proposal feels like homework. Building the thing feels like recess.

The trouble is that advice and a build are not the same kind of gift. Advice helps someone decide. A build is the product itself, handed over before anyone has agreed it's worth paying for.

My scorecard so far: 0 for 3

Last year I built a contractor a website he hadn't asked for. He never got around to looking at it.

In his defense, his family had an emergency that month. An unrequested website was never going to top his list. Looking back, I'm not sure it would have topped his list in a quiet month either. I had answered a question nobody asked.

In January I built three versions of a founder's app in under a day to impress him. It worked, sort of.

He looked at them and said I'd just proved how easy his app was to copy. I'd made his point for him, and for free. I sent a proposal the next day. It is still waiting for a reply, which seems fair. I had just shown him, in under a day, what I apparently thought the work was worth.

A couple of weeks ago I built a working trail app to show someone in my startup cohort what AI could do. He was so impressed that he got a Claude subscription and started building his own.

Technically, it was a success. Just not for my invoice.

If it helps, I'm in large company. The ANA and the 4A's surveyed agencies and clients for their 2023 "Cost of the Pitch" report. Nearly two-thirds of agencies (63%) said they deliver work free of charge during pitches. Only about a quarter of clients (26%) reported paying agencies anything for work done during a pitch.

So a whole industry runs on the same hopeful math I do.

When I build something in a day, the client sees a day

Here's what I missed. When I build something in a day, the client sees a day.

They don't see the almost four years I've spent building with AI tools. Or that the trail app's 90-second video came from a video setup I'd started in March for something else, and kept adding to in June.

From the outside, it looks like anyone with a Claude subscription could do it in an afternoon. Four years ago, it would have taken me a lot longer than an afternoon.

That gap is the whole problem with a free demo. The speed is real, but it's borrowed. It comes from years of trying tools, throwing most of them out, and building up a kit of pieces that happen to fit the next job. None of that shows up on screen. What shows up is a finished thing that apparently took one day.

So I've stopped thinking about this work in hours. AI shrinks the hours. Knowing what to build, and what to leave out, is the part that took four years.

To be clear, my rate card still has an hourly number on it. This is about how I think about the value of the work, not a new invoice format. But if the hours keep shrinking and the judgment doesn't, hours become a strange way to describe what a client is getting.

People value what they build themselves

There's even a name for what happened with the trail app: the IKEA effect. People value what they build themselves.

The name comes from a 2012 paper by Norton, Mochon and Ariely. A 2025 meta-analysis by Pelled and colleagues pooled the research and put the effect at d=0.57. For those of us who slept through that statistics class, that means it holds up, and it isn't small.

I showed him the finished bookshelf, and he went home and built his own.

I can't even be annoyed about it. He did exactly what the research says people do, and it was a good outcome for him. He's building now. I just shouldn't have expected the bookshelf to turn into a purchase order.

That's the part worth passing on to anyone who sells AI work. A free demo can teach the prospect that they can do it themselves. Sometimes that's the right outcome. It just isn't a sales strategy.

Fast builder, slow learner

The client this has gone best with is the one where, for once, I didn't build first. I gave advice for three weeks and waited for the retainer.

The retainer was his idea. His reasoning was that I should feel comfortable putting time into the work. Two weeks in, he started calling me his tech lead.

Nothing about that was faster. It was slower, and it's the one I'm still working on.

So my new rule, posted publicly so I can't quietly break it: my fastest work is for paying clients. If a prospect wants a demo, I offer a diagnosis instead. One call, then a written list of what I'd build and what it costs.

The diagnosis still gives something real away. The prospect walks out with a clear picture of the problem and a plan they could hand to anyone. What it doesn't give away is the build itself, which is the part they'd be paying for.

If you run a service business, the same rule probably travels. Give away the thinking that helps someone decide. Charge for the doing.

It took three free builds for that to sink in. Fast builder, slow learner.

What's the free work you'd take back?