Most "AI skills" lists for operations people are lists of tools.

Tools change every quarter. The skills that get you taken seriously by an owner don't.

I build AI systems for small businesses and I sit across from owners who have been burned by a chatbot or a consultant who couldn't explain what they sold. The people who get listened to, in my experience, share five skills. None of them is a prompt trick.

This is my opinion from that work, not a certification path. Here's the list, why each one matters and a simple way to practice it this month.

1. Mapping a process before touching a tool

If a team can't explain a process by hand, AI will only automate the confusion.

The most valuable thing an operations person brings to an AI project is the ability to write down how work really moves: who starts it, what triggers each step, where it waits, who finishes it and what "done" means. Not the version in the handbook. The version that happens on a Thursday afternoon.

Practice: pick one recurring process, like how an inbound quote request becomes a scheduled job. Write every step on one page. Then walk it with the person who does it and fix what you got wrong. If you can't get it on one page, that's your finding.

2. Choosing the right first project

Owners ask for AI to write content. The money is usually somewhere else.

For most service businesses, the best first project is close to revenue: missed calls that never get a callback, quotes nobody chases, past customers nobody asks back. I go deeper in my rule for what to automate first. The skill is picking the project by where money is leaking, not by what's easy to demo.

Practice: look at the last 30 leads. For each one, find out how fast the first reply went out and what happened next. Count where they dropped off. That list is your project backlog.

3. Setting a baseline and a check date

Without a baseline, nobody can say the thing worked.

Before anything is built, write down two or three numbers you expect to move, what they are today and the date you'll check them. Response time. Booked jobs per week. Hours spent on a task. This is the habit that separates an operations manager from someone who just likes new software, and it protects you when the owner asks, three months in, whether the project helped.

One caution that I come back to often: hours saved are not profit. Time saved only counts if it goes into something you can measure. I wrote about it in why hours saved is not profit and in AI Isn't Magic. It's Margin.

Practice: take any process you already run and write a one-line baseline for it. If you can't find the number, the first project is measuring.

4. Reviewing AI output like a manager

AI is confident whether it's right or not. Your job is to check it.

This is the skill I'd rank highest. AI reviews of long financial files can make confident mistakes: citing page numbers that don't exist and spending effort on conclusions that were already replaced. They can also find things a human missed. Both are true. The operator's skill is to treat the output as a smart opinion, supply the source documents and check the work.

In practice, that means:

  • Give it the documents. Don't ask it to remember. Paste or attach the source.
  • Ask it to show where in the source each claim comes from, then open the source.
  • Decide what needs human review. Anything that touches money or a customer gets reviewed until the system has a record.
  • Spot check forever. A weekly sample of ten outputs takes minutes.

Practice: take a task you know well, like summarizing a vendor contract or classifying a month of job notes. Run it through an AI tool, then grade the result against what you know. Keep a short log of where it was wrong. You'll learn its habits quickly.

5. Owning the system after launch

A tool is software. A system is software plus an owner plus a routine plus a check.

Most automations die quietly. A field changes in the CRM, a phone number gets updated, a form gets edited and the whole thing stops without anyone noticing. Operations people are the natural fit for this because their job is already about routines and follow-through.

What to own:

  • A named person for each automation.
  • A five minute weekly check: did it run, did it hit the right people, did anything error?
  • A short note on what it connects to, so the next person can fix it.

Practice: pick any automation your team already has, even a simple email rule. Find out who owns it and when it was last checked. Often the answer is nobody.

What you don't need

  • Coding. It helps. It's not required for the five skills above. Plenty of useful work happens in the tools owners already have.
  • A different tool every month. Pick one general-purpose assistant and one automation tool. Get good at them. A pile of subscriptions is a symptom, not a skill.
  • A certificate. If one helps you get a first conversation, fine. What owners respond to is a one-page process map and a baseline number.

How to use this in a real conversation

If you're an operations manager trying to be useful with AI inside your company, don't open with "we should use AI." Open with a finding. Here's the shape of it, with made-up numbers:

"Of the last 30 inbound leads, eleven got a first reply more than a day later. Here's the process as it works today. I'd like to fix this one thing, and here's how we'll know it worked."

That sentence uses all five skills, and it's the kind of thing that gets an owner's attention. It's also the starting point for deciding whether you need outside help at all. If you're weighing that, see when to skip an AI consultant.

If you'd like a second set of eyes on a first project, start with Di-Hy's free audit.

FAQ

What AI skills should an operations manager learn first?

Process mapping, choosing a first project near revenue and setting a baseline. These matter more than learning any specific tool, and they work with whatever tools your company uses.

Do operations managers need to learn to code to use AI?

No. Coding helps with some builds, but the highest-value skills are describing a process clearly, measuring results and reviewing output. Many useful systems run on tools owners already pay for.

How do I practice AI skills without a project at work?

Map one process on a page, pull a baseline from your last month of data, and run a task you know well through an AI tool, grading the output yourself.

Is a certification worth it for AI in operations?

It can help you start a conversation, but owners respond more to concrete work: a process map, a baseline number and a small improvement you can show.