How to pick your first automation.

The companies that get real value from automation rarely start with a grand AI strategy. They start with one well-chosen, slightly boring project — ship it, measure it, and let the result fund the next one. The hard part is the choosing. Here's the method we use.

Start with four questions

Sit down with the people who do the work — not just the org chart — and ask:

• Do we follow the same steps every week? Any process that runs on a rhythm — weekly reports, monthly invoicing, the Friday planning shuffle — is a candidate. Sameness is what automation feeds on.
• Is data being retyped from one system into another? Quote tool to invoicing, web form to CRM, spreadsheet to planning. Every retype is wasted time plus a chance for error.
• Do customers ask the same questions again and again? If your team can write the answer from memory, a well-fed assistant can draft it — and hand the unusual cases to a human.
• Is someone manually chasing people? Overdue invoices, unsigned contracts, quotes gone quiet. Reminders on a schedule are exactly the kind of work software doesn't resent.

A “yes” to any of these gives you a candidate. Most businesses end up with four or five. Don't do all of them — that's how automation projects die. Rank them.

Score on two axes

For each candidate, estimate two things. First, hours eaten: roughly how much time per week disappears into this task, across everyone who touches it. Second, mistake risk: if the automation gets one instance wrong, what does it cost? A mis-filed internal document is a shrug. A wrong number on a customer invoice is a phone call and a dent in trust.

Rough estimates are fine — you're ranking, not budgeting. Then pick from the high-hours, low-risk corner. That's your first automation: big enough to feel, safe enough that a teething problem is an annoyance rather than an incident. The high-hours, high-risk items aren't off the table — they're just not first. Do them once you've built trust in the tooling and in your own review process.

Ship one, and measure honestly

One automation, taken all the way to actually running, beats five prototypes. Before it goes live, write down what you expect it to save — hours per week, errors avoided — and then check, a few weeks in, against what really happened. Count the new costs too: reviewing the automation's output, handling its edge cases, the occasional fix. If the honest ledger is positive, take the next candidate off your ranked list. If it isn't, stop and understand why before automating anything else. Either way you've learned something real, which is worth more than a demo.

When not to automate

Honesty corner: sometimes the right amount of automation is none. Skip it when:

• The task is rare. Something that happens a few times a year won't repay the build, however annoying it is on the day.
• The process is still changing. Automating a workflow you're redesigning every month means rebuilding the automation every month. Let it settle first.
• Mistakes are expensive and hard to catch. If an error is costly and wouldn't be noticed until it's done damage, keep a human in charge — or automate only the drafting, never the sending.
• The real problem is the process. Automating a broken workflow gets you the same mess, faster. Fix the process, then automate the fixed version.

Sometimes plain software — or no software — beats AI. If that's your situation, the right advice is “don't”, and you should hear it before anyone bills you.

Want the ranked list done with you?

This exercise is precisely what our free AI Opportunity Audit is: a 30-minute call about how your business actually runs, then a written automation map within a week — what's worth automating first, roughly what it takes, and what we'd skip. If AI isn't worth it for you, the map says exactly that.

← Back to guides