I've been using an MCP — basically a plug-in that lets Claude talk to Klaviyo — to create and update emails for one of my clients. Most of the work, it does. Creates layouts and drafts. Pulls in segment logic. Builds the campaign and flows. Saves me hours. Literally, hours.

It's about 80% there.

The other 20% is the part where I'm in the loop. Cleaning up something that wasn't quite right. Uploading images. Rewriting a subject line that sounded like AI got close, but not quite. (I'm getting used to this by now.)

I think a lot of people in my position would look at that 80% and conclude the obvious thing — it's not ready yet, let's wait until it is. Fair. The ROI doesn't pencil cleanly. And honestly, most of the AI education people are consuming right now — the case studies, the content, the LinkedIn carousels — sets you up to expect 100%. Then you actually use the thing and it's like "oh, only 80%."

I'm not using AI because it works today

The bigger point here isn't "AI is good." It's a decision pattern. When something is directionally inevitable but operationally messy, I don't ask "is it perfect yet?" I ask: what decision today puts me on the right side of the learning curve — with the smallest cost and the biggest compounding upside?

I'm using it because I think it will, and the gap is closing fast. Faster than the conversation around AI is keeping up with.

The 80% I'm in right now isn't a problem I'm tolerating until the tools mature. It's the whole point. Every time I run a campaign through an MCP and watch it get most of the way there, I'm learning something I couldn't pick up by waiting — where it breaks, where it surprises me, what it just can't quite do, what a good prompt looks like for THIS specific job, what a bad one produces. That kind of fluency doesn't bootstrap. You don't catch up later by reading a guide.

By the time these tools are at 95% or 100%, the people who can actually use them well won't be the ones who waited for the press release. They'll be the ones who've already watched AI fail in interesting ways. They'll know which 5% they still need to own — and which 95% is finally cheap. That's a competitive position you can only build by being in the mess now.

What waiting actually costs

The way most leaders frame this is "we'll wait until AI is reliable enough to commit to." I get it. It's a reasonable position on paper.

But it's not really a wait — it's a hand-off of the learning curve. Someone is going to take it, and the cost of taking it is the same whether you do it now or in two years. The difference is just whether you take it while AI is the new thing, or after every competitor in your space has already taken it and you're scrambling.

The 80% isn't where AI is failing you. It's a low-cost window to learn how the tool actually behaves, before everyone else figures it out and the window closes.

The bet

I'd rather be wrong about AI now than right about it in two years. If I'm wrong — if it stalls at 80%, if the tooling never quite gets there — the worst that happens is I built up a sharper read on what AI can and can't do, which is a useful thing to know in any future where it matters. That's not a bad downside.

If I'm right — and I think the optimistic me is the one to listen to here — the people who got fluent in the messy middle are the ones who'll lead with AI when it's finally reliable. Everyone else will be three years behind, trying to figure out what the rest of us already learned by getting slightly wrong answers and improving them.

That's why I keep using AI even when it's only 80%. The 80% is the tuition.