FAQ Friday: “Is My Business Ready for AI Automation?”

Every week someone asks me a version of the same question: is my business actually ready for this? It’s the right question. It’s also the hardest one to get a straight answer to, because of who keeps answering it.

Yesterday my research agent ran that exact search — “is my business ready for AI automation.” Every result on page one was an agency. Nine out of nine, each attached to a readiness quiz that ends at a booking link. I sell automation builds too, so weigh this accordingly. The difference is I’ll show you my own homework, including the part where I fail.

Readiness isn’t about your tools

Most readiness checklists ask what software you own and how clean your data is. Useful, but it’s not the thing that decides whether automation sticks.

The real question is this: when the agent finishes its work, what has to happen next — and who does it?

Automation doesn’t delete work. It relocates it. Usually to a smaller, duller, much-easier-to-skip spot: approving a draft, pasting a reply, making a yes/no call. The build is the easy half. The half that kills projects is the human step nobody was assigned.

Here’s me failing my own test

One of my agents mines social threads for conversations worth joining and writes a tailored reply for each one. By design it runs in manual mode — it drafts, a human posts. That was my deliberate choice, because replying as me in someone else’s community is not a thing I wanted a bot doing unsupervised.

I opened the table this morning. Since 16 June it has written 299 drafts.

Posted: zero. The table that logs posted replies has 0 rows in it. Not a disappointing number — no number at all.

259 of those 299 drafts have already expired unread. That’s 86% of everything the agent produced, aged out because a thread goes cold after seven days. Forty are still alive as I write this; the oldest went into the queue on 10 September, and that first batch expires today.

Nothing broke. Every run finished clean, on schedule, for 94 days straight. The agent did precisely what I asked. The one step I kept for myself was never staffed, so the work quietly rotted at a rate of three a day while the dashboard stayed green.

That’s what “not ready” actually looks like. It doesn’t look like a failure. It looks like a system that’s working perfectly into a bucket with a hole in it.

The three questions that actually predict readiness

1. What’s the last step, and is it mine? Trace the workflow to the very end — not to “agent produces output,” but to the moment it’s live, sent, or paid. If the final step has your name on it, that step is the whole project.
2. If I do nothing for seven days, what happens to the output? Some queues wait patiently. Some expire. Mine expired, and I didn’t find out for three months. Know which kind you’re building before you build it.
3. Can I name the day and the person? Not “I’ll review it regularly” — an actual day, an actual name, even if the name is yours. If you can’t say it out loud, the loop is unstaffed and you already know how this ends.

Answer those three honestly and you’ll know more about your readiness than any quiz will tell you. Notice that none of them mention AI.

Today’s takeaway

You’re ready for AI automation when you can name the human step and the person who owns it. If you can’t, you aren’t blocked on technology — you’re about to commission a machine that fills a queue nobody empties.

There are only two honest fixes. Automate the last step properly so no human is required, which is what running something genuinely autonomous actually demands. Or keep the human step and put it on the calendar, the way I clear my approval queue every Sunday. What doesn’t work is the third option, the one I picked by accident: keep the human step and hope.

Still deciding where to point this first? Start with what to automate first — the answer is usually smaller than you think.

Want someone to trace your workflow to its real last step before you spend a cent on building it? Book an automation strategy session and we’ll map it together.

— Jon