Strategy · 6 min read
Why your AI plan should start with a list of things not to buy
Most AI budgets are spent before anyone has asked where the hours go. Here's the order that works.
The Human Office team
16 October 2026 · with Eloi Demeerseman
Every leadership team we meet has the same folder: a dozen AI tools someone tried, three subscriptions nobody cancelled, and one pilot that "showed promise". What's missing is a decision about what AI should actually do for the business.
The tool-first trap
Buying tools feels like progress. It's visible, it's fast, and it gives everyone something to talk about. But a tool chosen before the problem is defined will quietly shape the problem to fit itself.
“A tool chosen before the problem is defined will quietly shape the problem to fit itself.”
Start with the hours
Before any demo, ask each team one question: which tasks take hours every week and feel like they shouldn't? Rank the answers by time lost and by how often they repeat. That list is your real AI roadmap.
- Hours lost per week, per team
- How repetitive the task is
- What data the task needs, and who owns it
The not-to-buy list
Once the hours are visible, a second list writes itself: everything you were about to buy that doesn't touch those hours. In our experience, that list is longer than the first, and it pays for the whole exercise.
Key takeaways
- Define the problem in hours before you look at tools.
- Write the not-to-buy list next to the roadmap.
- Pick one use case per team, measured before and after.
What to do this week
Book one hour with your leadership team. Bring the list of AI tools you pay for. Leave with two lists: the three tasks worth fixing first, and the tools you'll stop paying for. That's a plan.