Automate a job when it’s the same every time, a mistake is cheap to fix, nobody on the other end cares whether a human did it, and you don’t need to get better at it. If any one of those fails, augment it: a person keeps the job and AI helps. Decide one job at a time.
If you’re a business owner, you’ve probably felt the fear already. AI is going to replace me, or replace my people, or replace the reason clients pick us. You don’t have to answer that for the whole business. You can answer it one job at a time.
What’s the difference between automating a job and augmenting it?
Automating hands a job to AI so it runs without you, while augmenting keeps you in the job and gives AI a supporting role, like reviewer or first-draft writer.
Both are useful. The mistake is picking one for the whole business. I’ve written before about what AI can’t do and why that’s the whole point, and about the fork every AI rollout hits. Below is a test you can run on a single job, before anything gets built.
Why does the automate-or-augment decision go wrong?
In my view, the decision goes wrong when the only picture of AI is AI doing the job instead of you. With only that picture, the answer to “should I use AI here?” is either yes or a flat no.
An executive coach I talked with recently gave me the flat no. He uses AI for admin and lead gen. He’d never use it for coaching. “It’s not genuine and authentic,” he said. Then: “I don’t know how you might prove me wrong.”
I agree people should coach people. But he was only thinking about it one way: AI coaching his clients, or AI replacing him.
So I offered a different job for the AI. Make it the reviewer. Like another person listening to your session and pointing out what you missed.
“Almost like someone to critique your performance,” he said.
Exactly. My suggestion was to record sessions 1 to 10, with the client’s OK, and ask the AI to look for patterns and changes over time. I told him it’s going to see other patterns he might have missed.
He hasn’t tried it yet.
How do you decide, job by job?
Pick one job in your business and answer four yes/no questions about it. A job is one unit of work: sending invoice reminders, writing a proposal, running a client session, answering a support email.
| Question about this one job | If yes | If no |
|---|---|---|
| 1. Is it the same every time? Same inputs, same steps, same right answer. | Automate candidate | Augment. It needs judgment. |
| 2. Do you want to be better at this job a year from now? | Augment. Use AI as a reviewer. | Automate, or hand it off. |
| 3. If it goes wrong, is it cheap and easy to fix? | Safe to automate | Keep a person on it. AI checks the work. |
| 4. Does the person on the other end care that a human did it? | Augment | Automate candidate |
Read it like this. Automate only when all four answers point to automate. If any answer points to augment, a person keeps the job and the AI gets a supporting role: reviewer, first-draft writer, pattern spotter.
Run a coaching session through it. It’s never the same twice. If you’re the coach, you want to get better at it. A bad session can’t be cheaply undone. And the client cares that a person is in the room. Four for four on augment. Then run the scheduling for that same session through the questions, and see where it lands.
Which jobs should you automate?
You should automate the jobs that pass the test. I’m not against automation. Plenty of jobs in a small business are good candidates.
Good candidates often include appointment confirmations, payment reminders, and moving data from a form into your CRM. Run each one through the table anyway. If it’s the same every time, cheap to fix, nobody on the other end cares whether you typed it, and you don’t need to get better at it, it passes.
I use the same reviewer idea I gave the coach on my own work. Several AI models review and check what I make before I review it myself. It saves a lot of time, and it catches multiple mistakes and issues regularly.
The danger is automating past the line. I said it in a video this week: “The easy thing is to automate stuff. The harder thing is to improve yourself.”
How do you run the test across your whole business?
You run it on a list of jobs, then group the results. Jobs bundle into seats, roughly the work one hire would have covered. Seats add up to the org, which is how your whole company runs once AI is inside it.
Start this week:
- Write down every recurring job you or your team did in the last two weeks.
- Run the four questions on each one.
- Automate one job that passed. Just one.
- Pick one job to augment, and set up an AI reviewer for it.
- Check back in a month. What did the reviewer catch that you would have missed?
If you’d rather do that mapping with someone, that’s what the AI Org Roadmap is. I interview you and audit how your business actually runs, then you get a written report with the process for building out an AI-first business. Start here.
Frequently Asked Questions
- What is the difference between AI automation and AI augmentation?
- Automation hands a job to AI so a person no longer does it. Augmentation keeps the person doing the job and uses AI to make them better at it, for example as a reviewer that spots patterns in their work. Most small businesses need both, decided one job at a time.
- How do I decide whether to automate or augment a job in my business?
- Check four things about that one job. Automate it only if it's the same every time, a mistake is cheap to fix, nobody on the other end cares whether a human did it, and you don't need to get better at it. If any one of those isn't true, augment it.
- Should a small business automate anything with AI?
- Yes. Jobs that run the same way every time, where a mistake is cheap to fix, are good candidates. Reminders, confirmations, and moving form data into a CRM are common examples to test. Automating those frees time for the jobs where your judgment is the product.
- What does using AI as a reviewer look like?
- You keep doing the work, then ask AI to review it. Record or save several versions of the same kind of work over time, then ask the AI to look for patterns and changes across them. It works like a second person listening in. It's going to see patterns you might have missed.
- Where should a small business owner start with AI?
- Start with a list of the jobs in your business, not a list of tools. Run the test on each job, automate the ones that pass, and pick one job to augment. Jobs group into seats, and seats add up to how your whole company runs with AI.