You bought the AI licenses. Now get people using them for actual work.
Six months after the rollout, the usage report shows a few heavy users and a long tail who opened the tool once. We run adoption programs that start from the work in each team, train people on their own tasks, recruit champions and measure usage, all under a policy people can follow.
Department leaderspriorities · sponsorsChampionsone or two per teamStaff by rolelegal · finance · service
02 · Program
Use-case portfolioranked · owner · data rulesRole-based trainingreal tasks · office hoursAcceptable-use policytools · data classes
03 · Approved tools
Workplace AI assistantCopilot · Claude · ChatGPTIdentity platformEntra ID groups
04 · Telemetry
Usage reportsadmin center · gateway logsAdoption dashboardrepeat use by task
Anthropic ClaudeOpenAI APIMicrosoft FoundryMicrosoft Entra ID
in short·20-second read
AI adoption is the work of getting employees to use AI tools well and safely in their daily jobs.
We run adoption programs with role-based training on real tasks, a ranked portfolio of use cases for each department and a network of internal champions. We track usage telemetry from tools like the Microsoft Copilot usage report. And we write a short acceptable-use policy that matches your data rules and the EU AI Act's AI literacy duty.
you probably need this if
The usage report shows a few heavy users and most licenses barely touched.
Staff ask in chat channels whether they're allowed to paste customer data into the AI tool.
Training was a single all-hands demo with generic prompts.
From the first call to a setup your own team can run.
Here's what usually prompts a call about your AI Adoption & Enablement, what we do about it and in which order, and where you end up.
01 · where it hurts02 · what we change03 · how it runs04 · where you end up
incident board · ai adoption & enablement3 open
openThe usage report shows a few heavy users and most licenses barely touched.
recurringStaff ask in chat channels whether they're allowed to paste customer data into the AI tool.
blockedTraining was a single all-hands demo with generic prompts.
exposure
Role-based training on real tasksUse-case portfolio by departmentChampions networkUsage telemetry that means…Acceptable-use policy people can…AI literacy evidence
startrehearsehandover
Baseline
Pick the first departments
Train and support
Measure and expand
we are here
ready to hand over
Usage starts showing up in Excel and Word.
The lawyers have a sanctioned tool and know what they can put in it.
Usage baseline by department, role and feature
Use-case portfolio per department with owners and data rules
chapter 01 · where it hurts
Why teams reach out
The usage report shows a few heavy users and most licenses barely touched.
Staff ask in chat channels whether they're allowed to paste customer data into the AI tool.
Training was a single all-hands demo with generic prompts.
chapter 02 · what we change
How we solve it
Role-based training on real tasks
Use-case portfolio by department
Champions network
Usage telemetry that means something
Acceptable-use policy people can follow
AI literacy evidence
chapter 03 · how it runs
Our approach
Baseline
Pick the first departments
Train and support
Measure and expand
chapter 04 · where you end up
What changes
Usage starts showing up in Excel and Word.
The lawyers have a sanctioned tool and know what they can put in it.
Usage baseline by department, role and feature
Use-case portfolio per department with owners and data rules
what we do
Where we spend the hours
6 areas of work, each one done by an engineer who has run it in production.
Sessions built around what a role actually does, such as a contracts lawyer comparing redlines or an analyst writing SQL, using your approved tools and sample documents. Prompt technique comes second to knowing which tasks to try.
Each team gets its own ranked list of use cases. We score them on frequency, time spent, data sensitivity and how easy the output is to check, and every entry says which tool, which data and who owns it.
One or two people per team get early access, deeper training and a direct line to the program team. In return they collect examples that worked, answer colleagues' questions and tell you where the policy gets in the way.
Active-user and feature-level data from the Microsoft Copilot usage report, the Copilot Dashboard in Viva Insights, or your AI gateway logs, joined to department and role. We track repeat use on the portfolio tasks. Logins alone say little.
A short policy that lists approved tools, which data classes may go into each, what needs human review, and how to request a new tool. Written with legal and security, then tested on real questions from staff.
We keep training records by role and tool, so you can show who was taught what about the systems they use. Article 4 of the EU AI Act asks providers and deployers to take measures for staff AI literacy, and this is the paper trail.
capability 01 / 06in production
Role-based training on real tasks
Sessions built around what a role actually does, such as a contracts lawyer comparing redlines or an analyst writing SQL, using your approved tools and sample documents. Prompt technique comes second to knowing which tasks to try.
Each team gets its own ranked list of use cases. We score them on frequency, time spent, data sensitivity and how easy the output is to check, and every entry says which tool, which data and who owns it.
One or two people per team get early access, deeper training and a direct line to the program team. In return they collect examples that worked, answer colleagues' questions and tell you where the policy gets in the way.
Active-user and feature-level data from the Microsoft Copilot usage report, the Copilot Dashboard in Viva Insights, or your AI gateway logs, joined to department and role. We track repeat use on the portfolio tasks. Logins alone say little.
A short policy that lists approved tools, which data classes may go into each, what needs human review, and how to request a new tool. Written with legal and security, then tested on real questions from staff.
We keep training records by role and tool, so you can show who was taught what about the systems they use. Article 4 of the EU AI Act asks providers and deployers to take measures for staff AI literacy, and this is the paper trail.
We pieced these together from AI Adoption & Enablement situations that come up a lot. They're patterns. You won't find a specific client in them.
case file 01
Finance has Copilot licenses and uses them mostly for meeting summaries
composite · not a client story
Beforethe situation
Finance got licenses for everyone at once. The usage report shows plenty of Teams meeting summaries. Excel and Word, where the team actually spends its day, barely register.
What we dothe work
Sit with the team through a close cycle and write down the tasks that come back every month, like variance commentary, reconciliation notes and policy lookups. Train each sub-team on three of them with its own files. One person in each sub-team becomes the champion.
Afterwhere you end up
Usage starts showing up in Excel and Word. Champions send back real examples, and the next session gets built out of those.
case file 02
Lawyers won't use AI until someone tells them what's allowed
composite · not a client story
Beforethe situation
Legal has access to an approved assistant, but nothing in writing covers privileged material. The general counsel's office says wait. Meanwhile some lawyers are using personal accounts.
What we dothe work
We wrote a one-page policy with the GC's office on privileged content and client data, and on when a human has to review output. Then we ran sessions on contract comparison and clause search with redacted matter documents.
Afterwhere you end up
The lawyers have a sanctioned tool and know what they can put in it. Personal accounts aren't needed any more. The GC's office holds everyone's training record.
How an AI adoption program connects training, usage and policy
Real tasks come from department leaders and champions and go into the use-case portfolio. Training by role on approved tools follows from it. Telemetry from the tools and the identity platform goes back to the program team, and they adjust portfolio and policy monthly.
01 · People
Department leaderspriorities · sponsors → Use-case portfolio · tasks
Championsone or two per team → Use-case portfolio · examples
Staff by rolelegal · finance · service → Workplace AI assistant · daily use
02 · Program
Use-case portfolioranked · owner · data rules → Role-based training · curriculum
Role-based trainingreal tasks · office hours → Staff by role · sessions
Acceptable-use policytools · data classes → Workplace AI assistant · data rules
03 · Approved tools
Workplace AI assistantCopilot · Claude · ChatGPT → Usage reports · activity
Identity platformEntra ID groups → Workplace AI assistant · license groups
04 · Telemetry
Usage reportsadmin center · gateway logs → Adoption dashboard · join to role
Adoption dashboardrepeat use by task → Use-case portfolio · monthly review
We don't start a step until the last one has produced something you can check.
step 01
Baseline
Pull current usage data, interview a sample of heavy and non-users, and review the existing policy and approved tool list.
step 02
Pick the first departments
Pick two or three teams with high-volume tasks and leaders who want it. Build their use-case portfolio and recruit their champions first.
step 03
Train and support
Training by role, with office hours afterwards. There's a shared prompt and example library, and a champion sits in each team.
step 04
Measure and expand
Monthly, we go through telemetry and what champions tell us, drop use cases that haven't stuck, and move the program on to the next departments.
you keep
What you're left with
We write these for your AI Adoption & Enablement setup, and they're working documents. Your team should be able to keep things running from them long after we've gone.
document01
Usage baseline by department, role and feature
yours to keep
document02
Use-case portfolio per department with owners and data rules
yours to keep
guide03
Role-based training materials using your tools and documents
yours to keep
document04
Champions network charter and shared example library
Some clients want a project team, some want a couple of engineers on theirs. Either works. Onsite, nearshore or offshore, with handovers across time zones that keep things moving.
Put these to anyone bidding on your AI Adoption & Enablement work. The ones who've done it answer differently from the ones who've read about it.
Q1
How will you tell real adoption from people opening the tool once?
Active-user counts look good after any announcement. Ask how repeat use on specific tasks will be measured and reported by team.
Q2
Who builds the training content, and will it use our documents and tools?
A generic prompt course rarely changes how anyone works. Train people on the tasks their role actually does, with the tools they're approved to use.
Q3
What will the policy say about each data class, and who signs it off?
People skip approved tools, or use unapproved ones, mostly because the data rules are unclear. Get legal and security to sign the same document.
Q4
What do champions get in return for their time?
Champion networks fade when the role is just unpaid extra work. Early access, some recognition and a direct line to the program team keep people in it.
Q5
What evidence of AI literacy training will we have at the end?
Deploy AI systems in the EU and Article 4 makes staff AI literacy an obligation. It's much easier to produce records by role and tool if you've kept them all along.
01Why is usage low when everyone has a license?
Usually because people weren't shown which of their own tasks the tool helps with, aren't sure what data they're allowed to paste in, or tried it once on the wrong task. Training on generic prompts doesn't fix any of those.
02What should we measure?
Weekly active users by department is the start. More useful is repeat use on the specific tasks in the portfolio, plus a short survey on time saved and output quality. Login counts alone reward curiosity, not adoption.
03Does the EU AI Act require AI training?
Article 4 has applied since 2 February 2025. Providers and deployers of AI systems must take measures so their staff have a sufficient level of AI literacy. Enforcement comes later. The Commission says the supervision and enforcement rules apply from 2 August 2026. Your counsel should confirm how this affects you.
04How long does an adoption program take?
Plan on a few months for a first wave in two or three departments. That includes a baseline, the training and two or so rounds of measurement. Going company-wide depends on how many departments and tools you include.
05Do you also build custom AI tools?
Yes. Our forward deployed engineers or generative AI team build whatever integration a use case needs. A lot of the portfolio needs nothing custom, though. A licensed tool and good teaching cover it.
06Should we block public AI tools while we roll out the approved ones?
Block it without offering an alternative and people move to personal devices. Most organizations get further by approving a tool, blocking only the riskiest categories and keeping the request process for new tools fast.
related
Often paired with
Projects on AI Adoption & Enablement usually end up touching one of these, so they're worth a look.
