Almost every AI conversation I'm in right now ends up in the same place. How many people can we replace.
And that's it. For a lot of companies, the whole ambition is to use AI to get smaller.
I get why. Cutting costs is easy to picture and easy to put in a spreadsheet. But it's half the equation, and it's the less interesting half. Because there's a second question almost nobody asks: what would we do with the time we just freed up?
There are only two ways to grow profit
Strip all the strategy decks down and profit moves in two directions. You either grow, by getting more customers or making each one worth more, or you cut costs. That's the whole menu.
I wrote a while back about where growth actually comes from, and the same four doors apply here. Sell more to the customers you already have. Understand them well enough to offer something new. Win new segments with what you've got. Or build something new for a new market.
Right now, though, almost everyone has turned to the cost side only. AI shows up and the first instinct is to reduce headcount. Understandable, short-sighted.
Here's the thing about the time you free up. It's raw material. Spend it on layoffs and you get a saving you can count, year after year, but it has a ceiling. You know roughly what the cut is worth and that's where it stops. Reinvest that same capacity into growth and there's no ceiling. It can pay back many times what the cut ever would, because growth keeps building on itself. Most companies still pick the cut, because it's the number that's easy to put in a spreadsheet. The bigger one, what that freed capacity could build, rarely makes it into the business case at all.
So the framing matters. You're not eliminating people, you're eliminating drudgery. The moment a transformation reads as code for layoffs, people stop helping it and start protecting themselves. After that it goes nowhere.
Nobody wakes up wanting AI
Nobody says "I'd love to buy some AI today." And it doesn't work like a plug you push into the wall that magically transforms everything either. What people have is a problem. They're frustrated, stressed, losing money. Something at work eats their time, their money, or both. That's what they actually want fixed.
In bigger companies especially, years of growth tend to leave fat behind. Layers, handoffs, work that stopped earning its keep a long time ago. So operational efficiency isn't the enemy here. A company has to keep an eye on it to stay in good enough shape to compete tomorrow. The mistake is treating it as the whole job and stopping there.
Which is why you don't start a transformation from the technology. You start from the business. Where is money leaking, where is growth stuck, where is the work just painful.
And yet I keep seeing companies hire AI engineers and technical roles before anyone has defined the business problem. That's backwards. Technical capability without business definition just builds the wrong things faster.
You can't redesign work you don't understand
This is bigger than automating a few tasks. It's redesigning how the work creates value, at every level: the role, the team, the whole organisation. The mechanism that turns effort into something a customer is willing to pay for.
First job: understand what the current work actually is. Scan for where repetition, slowness and errors hide. And here's the uncomfortable bit. Surprisingly often, the process isn't written down anywhere. Not even in large corporations. Everyone assumes they know how the work flows, but nobody has ever put it on paper.
That mapping is worth doing on its own, before any AI touches anything. It exposes the duplication, the unclear ownership, the points where work bounces from one team to the next and back again. Same friction I keep banging on about, it just hides better inside an org chart. And often the only person who actually sees and feels that friction is the customer. If it gets too high, if what they have to put up with starts to outweigh what they get back, you can guess what they do. They vote with their feet and go somewhere smoother.
Then there are two ways to redesign. The incremental one: find the spots in existing work where AI helps. Lower risk, faster. And the radical one: forget the current structure entirely. If you were setting this organisation up today, how would it run, how would it be led, what value does it create, for whom, how do you deliver it, and how does it capture that value? Most companies need the incremental version now and the clean-slate version as a direction to aim at.
Automation is the last step, not the first
This is the part most skip. You can't automate a process you haven't defined.
The sequence is understand, then map, then prioritise, then write down the actual logic, including the decisions and not just the steps, and only then automate. There's a simple test: if a human can't follow your notes, a machine can't either.
Skip this and you automate a broken process. Which means you've made the mess faster and more expensive. Growth scales friction, and so does automation. Excitement pulls everyone straight to building, and building is exactly the wrong place to start.
What's worth building
You can't do everything at once, so you prioritise. Plenty of people reach for ICE here, Impact, Confidence, Ease, and it's a fine first filter. Fast, gets people moving.
But watch what it does. Score on ease and confidence alone and your list fills up with cheap cost-cuts, because those always look easy. The framework quietly pushes you toward shrinking, the exact mistake we started with. So tag every opportunity: is this growth or cost? And put a readiness gate in front of the scoring. If the process isn't documented and stable, it's not an automation candidate yet, it's a documentation candidate.
The framework is secondary, honestly. The whole game is knowing what's worth building in the first place. The value never came from the model. It came from pointing it at the right thing. I've written before that the real bottleneck isn't building, it's deciding, and that's true here too.
And measure the right thing. Not how many tools you rolled out. Time saved plus income gained. Profit, not activity.
Someone has to own this
This work falls between IT and the business, and that gap is exactly where transformations stall. It needs an owner who understands operations and can speak technology. Business first, tech second.
If you're starting an AI transformation and you need someone to own the business side of it, to map the work, find where the value actually is, and lead it through, that's the work I do. Let's talk.
