Automate the Workflow, Not the Role: Why AI Rollouts Need a Chief of Staff and a COO

Every founder wants AI to make the company faster. The people best placed to make that happen well are the ones who already know how the work actually runs: the Chief of Staff and the COO.

I build with AI every week. I've automated invoicing, reporting, intake, and a good share of my own admin. So when a founder asks me "can AI just do this?", I'm not looking for a reason to say no.

The honest answer is usually: some of it. Which part depends on whether someone has taken the time to learn how the work gets done today. In most companies, that someone is the Chief of Staff or the COO. And I think that's quietly becoming one of the most important parts of both jobs.

A process is more than its output

From the outside, a process looks simple. The report goes out. The payment lands. The invoice gets sent.

From the inside, the person who runs it is doing things nobody wrote down. They compare this cycle to last cycle before anything goes out. They know which number always needs a second look. They catch the change that came in after the cutoff. They have a fallback for the day the data is wrong.

That's governance. It just doesn't look like governance, because it lives in a person's habits instead of a policy. A process that has run cleanly for years usually isn't clean by accident. Someone is keeping it that way.

Automating a workflow vs. replacing a role

Automating a workflow means you learn the process from the person who runs it, map every step, and ask why each one is there. Then you hand the repetitive pieces to AI: pulling data, formatting it, reconciling it, flagging what looks off. The checks stay. The person spends less time on assembly and more time on judgment.

Replacing a role means pointing a tool at the data and letting it produce the output. The visible part survives. The checks and fallbacks that lived in the person's head go with them.

The first one builds AI to support the work. Getting it right takes someone whose job is to understand how work moves through the company. That's where the Chief of Staff and the COO come in.

Where the Chief of Staff comes in: the map

A Chief of Staff sits between the leader and the work. They see every handoff. They know which meeting a decision really gets made in, who quietly fixes the report every month, and which team is carrying a process nobody else knows exists.

That makes the Chief of Staff the natural person to map a workflow before AI touches it. When the leader asks "can AI do this?", the Chief of Staff can come back with something better than yes or no. They come back with a map: these three steps can go to AI today, these two stay with a person, here's who owns each one, and here's what we'd watch for the first month.

In practice, that looks like:

1. Sitting with the person who runs the process for one full cycle.

2. Writing down every check, every exception, and every "I always look at this first."

3. Translating the leader's goal into a plan the team can actually run.

This is the same skill a good Chief of Staff already has. It's just pointed at a new question.

Where the COO comes in: the operating system

The COO owns how the company runs: how work moves, who's accountable, and what happens when something breaks. An AI agent inside a workflow is a new member of that system, and it needs the same things any new team member needs.

A named owner. Someone who answers for the output. A tool can't be accountable for a mistake. A person can.

Decision rights. Clear lines on what the automation decides on its own and what it routes to a person.

A plan B. The manual path still works. Someone knows how to roll back. The old and new versions run side by side for a few cycles before anyone retires the old one.

A pace. Start with a slice of the work and widen it as the track record grows.

This matters because automation doesn't usually fail loudly. I've written about the agent that fails quietly: wrong every cycle, at the same confidence, until someone downstream notices. Sometimes that's the next day. Sometimes it's months later, and in anything involving billing, refunds or customer funds, months of errors can become a compliance problem as well as a cleanup. The COO's job is to make sure there's a check in place to catch it early.

Four steps, and who owns each

1. Learn it from the person who runs it. The Chief of Staff leads. For every step, ask "what would happen if this didn't get done?" The answers are your controls.

2. Break it into pieces. The Chief of Staff and the process owner together. Gathering, calculating and formatting go to AI. Checks and decisions stay with a person, who now has better information in front of them.

3. Build plan B before plan A goes live. The COO owns it. Parallel runs, a working manual path, and a clear answer to "what do we do the day this breaks?"

4. Let trust be earned. The COO sets the pace. Expand what the automation handles as its record shows it's reliable, not because the demo looked good.

The person who built the process is your best asset

There's a people side to this that I think matters as much as the technical side.

These tools are new. We don't have much precedent for them yet. The person who has run a process reliably for years earned that trust one cycle at a time, and they know things about the work that no data export will show.

When they're brought in at the start, the rollout goes faster and lands better. They can tell you where the edge cases are. They become the person who checks the automation instead of the person worried about it. And the company keeps the judgment that made the process work in the first place.

That's what it looks like to build AI to support people instead of replacing them. Start by being genuinely interested in how they do the work.

The canary in the coal mine

There's one more part of both jobs that I think gets overlooked, and it may be the most important one.

It's easy to get caught up in AI. The excitement is real, the usefulness is real, and taking tedious work off people's plates is a good thing. But someone in the company has to be the gut check. Someone has to notice when the line moves from "AI is helping us" to "AI is doing the parts that should be human."

I think that's the Chief of Staff and the COO. The Chief of Staff hears how things land across the company before almost anyone else. The COO sees where the risk sits. Between them, they're the canary in the coal mine, and often the first ones in a position to say "we need governance here."

Here's where I see that line come up most.

Leadership communication. People can tell when an email was written by AI. When it comes from leadership, it lands as "they couldn't take five seconds to write to us." A note that's a little rough but clearly from a person does more for a team than a polished one that wasn't.

Sales and client relationships. Sales is built on relationships. Using AI to research a prospect or draft a first pass is smart. Sending a stream of automated emails that read like automated emails is eating your own dog food a little too far. The client can tell, and the relationship is what gets lost.

High-stakes systems. Some processes are so valuable, and so costly when they fail, that they shouldn't be handed to AI end to end. Anything that moves money, affects someone's livelihood, or carries legal weight keeps a person in it. AI can do the assembly. A human signs off.

The gut check comes down to a few questions:

1. Is AI supporting a person here, or replacing a conversation that should happen?

2. If this went wrong for six months, would anyone notice?

3. Would the people on the other end feel respected if they knew how this was done?

4. Are we still a company people can talk to?

That last one is the line in the sand. Efficiency is the point of AI. Becoming so dependent on it that the company forgets how to communicate like humans is not a trade worth making.

What this means for the roles

I've written before about the difference between a Chief of Staff and a COO. AI doesn't blur that line. It sharpens it.

The Chief of Staff becomes the person who understands how work really gets done and can turn a leader's "can AI do this?" into a plan. The COO becomes the person who makes sure that plan has an owner, a backup, and a pace. And both become the gut check on when it's too much. Together, they're how a company gets the speed of AI without inheriting problems it won't see until they're expensive.

Learn the workflow first. Then automate the pieces. Keep the people who know the work at the table.