Measuring and sustaining adoption
Adoption isn't done when the tool launches. This module covers how to measure, reinforce and scale it — and keep it current.
Learning objectives
- Measure reach, usage, proficiency, impact and experience
- Use feedback to adjust the programme
- Reinforce new ways of working
- Package and scale successful practices
6.1 Measure adoption — and impact
| Level | Example measures |
|---|---|
| Reach | % of staff with access and trained |
| Usage | Weekly active users, frequency, breadth of tasks |
| Proficiency | Confidence surveys, quality of outputs, use of advanced features |
| Impact | Time per task, quality, customer satisfaction, business outcomes |
| Experience | Employee sentiment, workload, concerns raised |
Usage alone can mislead: people may use AI a lot for low-value tasks. Combine usage with impact and experience.
6.2 Listen and adjust
- Run short pulse surveys every few months
- Hold retrospectives with pilot teams: what helped, what got in the way
- Track questions and issues raised with champions
- Look for teams with low adoption and ask why — the answer may be a real problem with the tool or process
6.3 Reinforce the change
The R in ADKAR — reinforcement — is where many programmes stop too early. Sustain change by recognising people who find valuable uses, building AI into onboarding, job descriptions, processes and performance conversations, sharing success stories regularly, and keeping leaders visibly involved.
6.4 Scaling what works
When a team finds a valuable way of working, package it — the process, prompts, training and measures — so other teams can adopt it quickly. Champions and communities of practice spread it; the central AI team removes obstacles and keeps quality and risk controls consistent.
6.5 Keep learning
AI tools change every few months. Treat adoption as a continuous capability, not a one-off project: refresh training, update policies and prompt libraries, revisit workflows when tools improve, and keep listening. Free, reputable training from Anthropic, OpenAI, Microsoft, AWS, Google and IBM — linked throughout this site — makes it easy to keep skills current.
Design a simple adoption dashboard for your area with one or two measures for each of the five levels in 6.1, a target for each, and how often you'll review it.
See a worked example
Example: Reach — 95% trained (target by Q2). Usage — 70% weekly active. Proficiency — 75% 'confident' in pulse survey. Impact — report drafting time down 30%; error rate not increased. Experience — workload rating stable or better; concerns raised and resolved count. Reviewed monthly by the leadership team, quarterly with champions.
Key takeaways
- Combine usage with impact and experience.
- Ask why adoption is low — it may reveal a real problem.
- Reinforcement sustains change: recognition, onboarding, processes.
- Package what works so others can adopt it.
- Treat AI adoption as a continuous capability.