Companies over 100 staff are being pitched a copilot every week[1]. Some land. Most do not. Here is the pattern of which.
Adoption curves
Source: My client analytics, anonymised
Three client deployments tracked over 12 months. Company A had 79 percent monthly active rate at month 12, a real productivity tool. Company B faded to 26 percent, still used but only by a third of staff. Company C dropped to 5 percent, effectively dead.
The difference is workflow fit, not model quality. All three used the same model.
Five questions
| Spec | Question | Strong yes → | Weak → |
|---|---|---|---|
| Specific task it replaces? | Yes (replace 30+h/mo manual) | No (chat interface) | |
| Measurable output? | Yes (rows in CRM, fewer tickets) | No (vibes) | |
| Workflow integration? | Lives in existing tool | Separate chat tab | |
| Cost transparent? | £X per result | "AI is bundled" | |
| Owner can debug? | Logs visible | Black box |
These five questions predict success in advance. If any are weak, the deployment will struggle.
The single biggest determinant: does the copilot replace a specific task that someone does for hours per week? If yes, adoption sticks. If it is a generic "ask me anything" interface bolted on, adoption fades.
Concrete examples
Worked: Customer support copilot that drafts ticket responses inside Zendesk. Saves 30-40 minutes per agent per day. Used 80 percent of working hours.
Worked: Sales copilot that drafts follow-up emails inside Salesforce. Saves 2-3 hours/week per rep. Used daily.
Did not work: Generic Copilot in Microsoft 365 deployed to all staff. After three months, only the marketing team and a handful of execs used it weekly. The licence cost was visible; the productivity was not.
My advice to clients
Pick one workflow. Replace one task. Measure the time saved. If it works, expand. If it does not, kill it.
Avoid blanket licence rollouts. Avoid "AI is bundled" deals. Pay for what you use, on the workflow that uses it.
The technology is good enough in 2026. The challenge is workflow design, not model selection.
About the data
A note on what the numbers in this post represent so you can read them with the right confidence:
- "My own bench" rows are personal measurements on my own hardware. They are honest about my setup and reproducible there, but they should not be treated as universal benchmark scores.
- Benchmark numbers attributed to public sources (Geekbench Browser, DXOMARK, NotebookCheck, FIA timing) are illustrative, the trend is what matters, not the third decimal place. Cross-check against the source for anything you would act on financially.
- Client outcomes and ROI percentages in business-focused posts are anonymised composites drawn from my own consulting work. Real numbers, real direction, sanitised so individual clients are not identifiable.
- Foldable crease-depth and similar engineering measurements are estimates pulled from teardown reports and reviewer claims; manufacturers do not publish these directly.
- Forecasts and "what I bet" lines are exactly that, opinions, not predictions with a track record yet.
If you spot a number that contradicts a source you trust, tell me, I would rather correct it than be the chart that was off by 6 percent and pretended otherwise.