How I put a Claude Managed Agent to work

Most AI projects in a small business fail the same way. Someone buys a tool, it sits half-used, and three months later nothing about the actual work has changed except a new monthly subscription. When I put a Claude Managed Agent into a business, I work in the opposite direction. I spend almost none of my time on infrastructure and almost all of it on your work — what you do every week, where the time goes, and which slice of it an agent can quietly take off your plate.

Here is the method, start to finish.

Start with the work, not the tech

I don't open with technology. I sit with you and map how a job actually gets done today, end to end — every spreadsheet, every copy-paste, every hand it passes through before it's finished. You can't automate a workflow you haven't watched.

So the first questions are not about AI at all. How does a job arrive? How many hands touch it before it's done? Where does the same information get re-typed into a second system? What's the bit everyone quietly dreads on a Monday? I'm mapping the real flow, not the tidy version on the org chart — because the gap between those two is usually exactly where the time is leaking.

Why I build a thin layer, not a platform

Claude's Managed Agents already own the hard parts — running the agent, isolating it safely, holding your credentials, and keeping a durable record of every session it runs. There's no database for me to build and no server for me to babysit. I put a thin layer on top of that, and nothing more.

That choice is the whole trick. Because I'm not rebuilding plumbing that already exists, the build is small and the risk is low — and my time goes where it actually matters.

The platform owns the plumbing. That is what frees me to spend my time on your business instead of mine.

Do you actually need an agent?

This is the honest gate, and I run it before anyone spends money. Sometimes a subscription is enough. Sometimes a scheduled script is enough. An agent only earns its place when the work has a human approval step, has to run without you babysitting it, or has to produce the same output for several people every time.

If a job is a one-off, or one technical person can already do it in a chat window, it doesn't need an agent — and I'll tell you that. I'd rather lose the sale than wire up something you'll be paying for and not using by spring.

The real work: skills, tools, context, routines

This is where almost all of my time goes. The platform gives you a capable worker out of the box, but a capable worker who knows nothing about your business is no use. The value is in what you teach it.

Get those four right for one real job and you have something genuinely useful. Get them wrong and you have a clever demo that nobody trusts with real work.

A human stays in the loop

The agent does the first pass. Nothing client-facing or financial goes out until a person has approved it — the agent prepares, you click. That single gate is what makes the whole thing safe to put into a real business, and it's cheap insurance for the first few months while you build trust in the output.

And it lives where you already work. Not in yet another app you have to remember to open — in the inbox, the chat, or the dashboard your team is already in every day.

Start with one seam, then compound

I start with one narrow seam of your workflow — a single step the agent owns end to end — prove it for a couple of weeks, then expand. A seam is easier to ship and far easier to roll back than a full rebuild, so the downside if it doesn't work is small.

I price it against what the work is worth to you — a line on a salary, not a token meter. The compute behind a small stack of agents is a rounding error; the value is the hours it gives back. Land one workflow that runs reliably across the whole team, and that's worth more than ten that each only half-work.

If you want to see where an agent earns its place in your business — and, just as importantly, where it doesn't yet — that's what a short AI readiness audit is for. The first step is always a plain conversation about what you actually do every week and where the time goes.