A founder I worked with last year had four active AI subscriptions. ChatGPT, Claude, Notion AI, and a specialist tool for their industry. Total spend: around €400 a month. Usage across the team: sporadic. ROI: close to zero.

When I asked which business problem each tool was solving, there was a long pause. "We thought they'd figure it out," she said. "The tools are so capable — we assumed the team would find the uses."

This is the founder's mistake. And it's remarkably common.

Where most teams are when they call me
💸
High spend, low clarity
Multiple subscriptions active. No defined use cases. Team using tools ad hoc or not at all.
🔍
One tool, unclear fit
Committed to a platform. Not sure it's the right one. Team using it for low-value tasks.
🎯
Problem-first, then tool
This is where you want to be. Tool chosen because it solves a specific, named problem.

The instinct isn't wrong — the order is

Founders who invest in AI tools early are doing the right thing. The instinct to equip the team, to stay ahead, to not get left behind — that's sound thinking. The mistake isn't the investment. It's the sequence.

Most teams buy first and ask questions later. They see a compelling demo, read about what competitors are doing, feel the pressure to modernise — and they subscribe. Then they try to retrofit the tool onto their existing workflows and wonder why it doesn't stick.

The teams that get real ROI from AI do it in the opposite order. They map the problem first — specifically, they identify the tasks that are eating the most time with the least payoff — and then they find the tool that fits those tasks. That inversion changes everything.

"The question isn't 'what can this tool do?' It's 'what is currently costing us the most time, and is there a tool that solves exactly that?'"

The right order of operations

1
Map the bottlenecks
Spend half a day with the team listing every repetitive task that happens more than twice a week. Don't judge. Just list.
2
Score by time and pain
Which tasks take longest? Which ones frustrate people most? The intersection of slow and painful is where automation pays off fastest.
3
Pick one task to start
Not three. Not five. One. Solve it properly before moving to the next. A single visible win builds more momentum than five half-finished automations.
4
Now choose the tool
With a specific task defined, the tool choice becomes obvious — or at least much narrower. You're no longer evaluating features. You're evaluating fit.
5
Build and hand over
Deploy the solution, document it, and train the people who'll use it. Then repeat with the next bottleneck.

What buying first actually costs you

The subscription fees are usually the smallest part of the cost. What gets lost is harder to measure but much more significant.

🕐
Team time
Every hour spent exploring a tool that wasn't designed for your problem is an hour not spent on work. Multiply by team size.
😤
Goodwill
Teams that try a tool and get nothing from it develop resistance. The next rollout — even a better one — faces a credibility deficit before it starts.
📉
Momentum
Failed AI initiatives create inertia. The question stops being "what should we automate next?" and becomes "should we even bother?"

What to do if you've already bought

If you're already subscribed to tools that aren't delivering, you don't need to cancel and start over. You need to do the problem-mapping exercise you should have done first — and then figure out which of your existing tools, if any, fits the problems you've identified.

Sometimes the tool you have is right and just needs to be deployed differently. Sometimes you're using a Swiss Army knife for a job that needs a scalpel. Either way, the answer starts with getting clear on the problem — not the tool.

That half-day audit is often the most valuable thing I do with a new client. It costs nothing except time, and it tends to completely reframe what the next six months look like.