What I Would Automate First in a 5–50 Person Company

A founder I spoke with last spring had just bought a sales automation suite. Eleven seats, annual contract, a logo I recognized from airport ads. He showed me the dashboard with some pride. Then his phone buzzed, and he excused himself to approve a delivery note — by hand, from his phone, in the car park — because the warehouse couldn’t release the shipment without his initials. He’d been doing that four or five times a day for years. The expensive thing automated something he did once a week. The thing he did five times a day, every day, he’d never thought to touch.

That’s the pattern. Owners almost always automate the wrong thing first.

The wrong thing is usually the loud thing

The wrong thing is loud. It has a category, a competitor everyone’s heard of, a salesman who returns your calls. It feels like progress because it came with an onboarding call and a Slack channel. It announces itself.

The right thing is quiet. It doesn’t have a category. Nobody’s selling it to you, because it’s shaped exactly like your business and nobody else’s. It’s the delivery note in the car park. It’s the spreadsheet your office manager re-keys every Friday from one system into another. It’s the “let me just check” that interrupts you eleven times a day.

That’s not the one you’ll be sold. The boring one is.

Three things have to be true

The first task to automate isn’t the most expensive task, or the most visible, or the most modern. It’s the one where three things are true at once.

It’s repetitive — it happens often enough that the time compounds. Not the big quarterly reconciliation. The small daily thing nobody counts because each instance is only four minutes.

It’s error-prone — it’s done by hand, from memory, under interruption, which means it goes wrong often enough that someone’s always cleaning up after it. A wrong figure typed into the second system. A delivery note approved for the wrong client because you were in a car park.

And it sits on you — the owner — as a bottleneck. Work stops until you touch it. The shipment waits for your initials. The quote waits for your number. You are the single point of failure for a thing that requires none of your judgment.

When all three are true, you’ve found it. Not before.

How to actually find it

You won’t find it by listing your processes. You’ll find it by watching where your day breaks.

Follow the interruptions. For one week, notice every time someone stops what they’re doing to come ask you something. Most of those questions have the same answer every time. That sameness is the tell.

Follow the phrase “let me just check.” Every time you or someone on your team says it, write down what they checked and where. A status, a stock level, whether an invoice went out. “Let me just check” is the sound of information that should be visible but isn’t, being looked up by hand.

And follow the things done from memory. The order someone “just knows,” the supplier email someone “always sends,” the step that lives in one person’s head and nowhere else. Anything carried in a head is a thing that breaks the day that head is on holiday.

Those three trails — the interruptions, the checks, the things done from memory — meet at the task that deserves to go first.

What to leave alone

Not everything that’s annoying should be automated. Some work is annoying because it’s supposed to involve a person.

Leave the judgment calls alone. Pricing a job that’s never been priced. Deciding whether to extend a deadline for a client who’s earned it. The moment you systematize a judgment call, you get a fast, confident, consistently wrong answer.

Leave the relationship alone. The call to the supplier when something’s gone wrong. The note to the client who’s gone quiet. Automating the relationship doesn’t save time — it spends trust you can’t easily earn back.

And leave the twice-a-year thing alone. The work you do rarely is rarely worth the cost of automating. You’ll spend two weeks building something to save yourself an afternoon in June and another in December. Do it by hand. Write down how, so the next person can too. That’s enough.

The reframe

Here’s the part worth slowing down for. Automation isn’t about doing more, faster. That framing is how owners end up with a faster version of a broken process — eleven seats of speed pointed at the wrong work.

Automation is about removing the work that should never have needed a human in the first place.

The delivery note didn’t need the owner. It needed a rule: this client, this amount, release it. The Friday re-keying didn’t need a person. It needed the two systems to talk. None of that is about speed. It’s about taking the human out of the place a human was only ever a bottleneck — and putting that human back where judgment, relationship, and consequence actually live.

That’s the whole filter. Repetitive, error-prone, sitting on your desk — that goes. Judgment, relationship, the twice-a-year thing — that stays.

The founder with eleven seats never did fix the delivery note while I was watching. Last I heard, he was still approving them from the car park. The dashboard’s very nice. But the thing he does five times a day, every day, is still waiting for a question nobody on the sales call thought to ask: what’s the work that should never have needed you at all?

That’s the one I’d start with. If you’re not sure which task it is in your business, that’s exactly what a Diagnostic is for.