Driving AI adoption vs. buying more licences: which actually pays back?
You can buy more Copilot seats, or get people using the ones you already have. With adoption stalled near 5%, more licences rarely pay back. Here is how the two approaches compare.
Short answer: with most AI licences already going unused, buying more seats almost never pays back — driving adoption of what you already own does. Adoption recovers spend you have already committed; new licences add cost on top of a gap you have not closed yet.
The two approaches, side by side
Buying more licences assumes the constraint is access — that more people with the tool means more value. Driving adoption assumes the constraint is behaviour — that value comes from people actually using the tool on real work. The evidence favours the second: Microsoft 365 Copilot adoption has stalled around 4.5% with weekly usage near 1%, and fewer than 4 in 10 licensed users are active. Adding seats to that does not change the ratio.
Where buying more licences makes sense
- You have measured utilisation and existing seats are genuinely maxed out.
- A specific team has a proven workflow and needs to scale it to more people.
- The tool is already adopted and the return per seat is established.
In other words: buy more only *after* adoption is working, not as a substitute for it.
Where driving adoption wins
- Utilisation is unknown or low (the usual case).
- Licences were bought ahead of a rollout that never fully landed.
- People are quietly working around the official tools (shadow AI).
- You need the spend to show a return this quarter, not next budget cycle.
Adoption is self-financing in a way new licences never are: cutting idle seats frees the budget that funds the enablement, so the return does not require fresh spend.
The cost comparison that matters
More licences: a known, recurring cost with an unknown return. Adoption work: a one-time investment, funded from recovered waste, that raises the return on every seat you already pay for. When only ~16% of AI pilots reach production (Gartner data on generative AI shows ~6% of enterprises get GenAI to production at all), the leverage is clearly on adoption.
Common questions
Isn't more seats the faster path to impact? Only if the current seats are fully used. If they are not — and half of SaaS licences sit idle industry-wide — more seats just scale the waste.
Can we do both? Yes, in order. Drive adoption first, measure the real return per seat, then scale licences into proven demand.
How do we know which we need? Measure utilisation. If usage is low, the problem is adoption and no amount of new licences fixes it.
Licences are access. Adoption is value. You have already paid for the access — the return comes from closing the gap, not widening it.