It’s the most common question I get on discovery calls, and it’s a fair one. You’re paying for ChatGPT or Claude, maybe a meeting note-taker and a couple of automations. Why pay a person to run operations when the software is getting better every month?
The short answer: AI is very good at doing work, and not good at owning it. If your problem is that tasks take too long, AI may be all you need. If your problem is that things fall through the cracks, decisions wait for you, and nobody is watching the whole business, a tool won’t fix that. Let me explain the difference, and how to work out which problem you actually have.
What AI genuinely does well in operations
I use AI every day in client work, so I’m not here to talk it down. These are the jobs where it saves real time:
- First drafts. SOPs, client emails, proposals, job descriptions, meeting agendas. A decent prompt gets you 70% of the way in seconds.
- Summarising. Long email threads, call transcripts, contractor updates and reports, turned into a few lines you can act on.
- Sorting and tagging. Reading incoming enquiries and labelling them by type, urgency or service.
- Turning messy notes into structure. A voice note becomes a task list. A rambling brief becomes a checklist.
- Repetitive handoffs, when it’s paired with an automation tool like Zapier or Make. A form comes in, AI drafts the reply, the CRM updates, and a task is created.
If most of your week goes on these jobs, start by learning to do them well with AI. I’d genuinely rather you didn’t hire anyone for them.
What AI can’t do, however good the prompt
Here’s where it stops. None of these are about the model being clever enough. They’re about responsibility.
1. It doesn’t decide what matters this week
AI answers the question you ask it. It doesn’t notice that your biggest client hasn’t renewed, that a contractor has gone quiet, or that three small delays are about to become one missed launch. Setting priorities means weighing things against each other with context only someone inside the business has.
2. It doesn’t check its own work reliably
AI output is confident even when it’s wrong: a made-up policy in an SOP, a wrong date in a client email, a summary that misses the one sentence that mattered. Someone has to review what goes out, and that someone has to know the business well enough to spot the error.
3. It doesn’t follow through
A tool won’t chase the invoice that’s 14 days late, nudge the designer who missed a deadline, or notice that the onboarding email never fired because a form field changed. Follow-through is noticing what didn’t happen, and AI only ever responds to what did.
4. It doesn’t maintain itself
Your business changes. You add a service, raise prices, hire a contractor, switch tools. Every change quietly breaks a prompt, an automation or an SOP somewhere. Without someone tending the system, AI workflows decay. After six months, most of them are either switched off or doing something slightly wrong.
5. It can’t be accountable
When something breaks, a client is upset or money is lost, someone has to answer for it and fix it. That’s the real job of an operations manager, and it can’t be handed to software.
A quick test: tools problem or ownership problem?
Answer these honestly.
- Could you take a full week off without checking messages, and come back to a business that ran normally?
- Do you know, right now, which leads are waiting for a reply and for how long?
- If a contractor missed a deadline yesterday, would you find out before the client did?
- Are the AI tools you pay for connected to each other, or do you copy and paste between them?
- When did someone last review whether your automations still do what they were built to do?
If you answered “no” or “I’m not sure” to two or more, your problem isn’t speed. It’s that nobody owns how the business runs. More AI will make you faster at the work, but it won’t stop things slipping.
What an operations manager actually does in an AI-heavy business
The role has changed. Five years ago, much of an OBM’s week went on execution: scheduling, inbox triage, drafting documents. Today a good operations partner spends that time on the layer above:
- Running the weekly rhythm. A Monday priorities list, a weekly metrics review and a monthly look at what’s working. You get clarity without having to chase it.
- Designing hybrid workflows. Deciding which steps AI handles, which a person handles, and where the human check sits before anything reaches a client.
- Quality control. Spot-checking AI output, contractor work and automations, the same way you’d review a new hire.
- Tool upkeep. Updating prompts, fixing broken automations and cutting the subscriptions nobody uses.
- Managing people and vendors. Briefing contractors, chasing deliverables and handling handoffs.
- Catching what slips, and answering for it when something goes wrong.
That’s an operator, not an assistant, and it’s the part that’s hardest to automate.
The cost question
The obvious objection: “ChatGPT costs $20 a month. You cost a lot more than that.”
Compare it to the right thing. An operations partner isn’t competing with a subscription. They’re competing with:
- Your own hours. If you spend ten hours a week managing tools, people and loose ends, value that time honestly. At even $100 an hour, it’s about $4,000 a month.
- Leakage. One lead that went cold, one renewal nobody chased, one client who left because onboarding felt chaotic.
- Failed experiments. The AI tools you bought, set up halfway and abandoned.
- The hire you’re about to make. Good systems often let you delay a full-time hire by months.
If someone gives you back ten hours a week and stops one client slipping away, they’ve usually covered their fee before you count anything else.
When you don’t need one
To be fair, here’s when I’d tell you not to hire an operations manager:
- You’re a solo business with simple delivery and one or two revenue streams.
- Your main bottleneck is getting clients, not serving them.
- You enjoy running the back end and you have the time for it.
In those cases, learn a few AI workflows well, set up one project tool properly, and revisit the question when you add your first contractor.
Where to start
If you’re not sure which camp you’re in, start with an audit rather than a commitment. Map how work actually moves through your business, from lead to delivery to payment. List every tool and AI subscription you pay for. Note where you personally are the bottleneck. That alone will tell you whether you need better prompts or a person who owns the system.
If you’d like me to do that mapping with you, that’s exactly what my Operations & AI Audit covers. It’s two weeks, fixed price, and the plan is yours to keep either way.