An AI adoption programme: licences into habits.
For teams that bought Claude or ChatGPT and found that only a few people use it. We work two workflows at a time: timed in week one, built into skills on your real work, taught in the room, and timed again.
The short answer
Teams use Claude or ChatGPT when each person has a named job to do with it, not when they have a licence. An AI adoption programme picks two recurring workflows at a time, times them in week one, builds each into a skill with the person who does the job, teaches it in the room, and times it again. Sentry AI runs this for teams in New Zealand and Australia.
Who the programme suits
Teams of roughly five to fifty people that have bought Claude or ChatGPT, or are about to, with a few enthusiasts and a lot of quiet seats. Owners who want to know in hours, not opinions, whether it worked.
Not a fit: one person learning on their own, where a free self-paced course is the better start, or a team that needs a certificate for every attendee.
The programme
Five steps, two workflows at a time
The first round runs over the first month. Each round after it adds the next two workflows, so the habit spreads job by job instead of seat by seat.
- 01 · Week one
Name the goals and time the work
Up to three goals, and the jobs behind them timed as they run today. That baseline is what every result is measured against, in a unit your team chooses: minutes a week, hours a month.
- 02 · Week one
Pick two workflows, not twenty
Two recurring jobs that eat real time, chosen with the people who do them. Not the most impressive ones. Two is few enough to finish properly and enough for the rest of the team to see it working.
- 03 · The training day
Build them into skills, taught in the room
Each workflow built into a Claude or ChatGPT skill with the person who does the job, on their own work, and taught where they work. Everyone set up, with a one-page policy on what may go in.
What the training day covers - 04 · End of the first month
Time it again
The same jobs, timed the same way, against the week-one baseline. If a workflow is not faster we keep working on it, and if it was the wrong pick we say so and choose another.
- 05 · Every round after
The next two, and a daily eval
Two more workflows per round, each with its own baseline. Between rounds, a daily read of what the team actually did with its skills proposes the next fix, and a named Sentry AI engineer ships it with you.
The method behind the programme
What is included in the first round
A goals register
Up to three goals, each with a week-one baseline and the unit it is reported in.
Two working skills
Built on your real work, owned by your organisation, and running for the people who use them.
A training day in the room
Everyone on your Claude or ChatGPT organisation, with a short policy on what goes in.
Your business written down
The context a model cannot know: products, terms, tone, how a job is done here.
The re-timed result
Each workflow measured again against its baseline and reported as a number you can check.
The adoption scorecard
Adoption or activity? Four numbers to report instead
Most AI dashboards count activity. A team can send thousands of prompts and save no time at all. These are the four numbers we report, and the ones they replace.
| Counts activity | Counts adoption |
|---|---|
| Seats active this month | Workflows named, with a person who owns each one |
| Prompts sent | Workflows with a week-one baseline |
| People who opened it | Workflows re-timed and faster than their baseline |
| Training sessions attended | People with a skill running on their own work |
AI adoption, AI enablement or change management?
Three different things are sold under the same words. Pick by where each one starts.
| Our adoption programme | Change-management programme | Licences plus an online course | |
|---|---|---|---|
| Starts with | Two named workflows and a baseline | A communication and training plan | Logins |
| Where people learn | In the room, on their own work | Workshops and champions | Online, alone |
| What changes first | Time on a job they already do | Awareness and attitudes | Feature knowledge |
| How you know it worked | Each workflow re-timed against its baseline | Surveys and usage dashboards | Course completion |
Teams we have moved over
Measured against their own baseline
A five-person ecommerce team moved its core tasks from about four hours to about one against its own baseline. A retailer runs seven departments through Claude.
The programme is run by the people who build the skills it teaches. We run our own business on Claude, and keep building and maintaining skills for our clients every day.
Every answer, skill and training day comes from James Oldham, who founded Sentry AI in Auckland and does the engineering on the work these pages describe. About James
- people connected to the AI systems we run
- 185+
- skill runs a month across client teams
- 1,500+
- skills in a single client's library
- 45
- organisations we run AI for today
- 7
Live figures from the teams we work with, October 2026.
AI adoption questions
Why does nobody use the AI licences we bought?
Because a licence gives people a tool and no job to do with it. Most teams end up with a few enthusiasts and a lot of quiet seats. Adoption starts when each person has a named job they repeat every week, built into Claude or ChatGPT on their own work, and taught in the room where they do it.
What is an AI adoption programme?
A structured way to move a team from owning AI licences to using them on real work every week. Ours works two workflows at a time: we time each job in week one, build it into a skill with the person who does it, teach it in the room, and time it again. Then we pick the next two.
Is this the same as AI enablement or AI change management?
It covers the same ground in a different order. Change-management programmes usually start with communication and training plans. We start with the work itself: two named workflows, a baseline, and a skill built on them. The behaviour change follows from people getting time back on a job they already do.
How long does an AI adoption programme take?
The first round runs over the first month: goals and baselines in week one, a training day in the room, the first two workflows built and re-timed by the end of the month. Each round after that adds the next two workflows. Most teams can see in the first month whether it is working, because the numbers are their own.
Does it work for ChatGPT as well as Claude?
Yes. We start with whichever your team already pays for. Most of our teams run on Claude, and we will say plainly if a licence you hold is not worth renewing.
How do you measure AI adoption?
Not by logins, prompts sent or seats active: those count activity. We report how many workflows are named, how many have a baseline, how many have been re-timed faster, and how many people have a skill running on their own work. Hours are reported in the unit the team chose in week one, such as minutes a week.
Do you guarantee a result?
No. We set a goal for each workflow against its own baseline, by default twice as fast, and report the measured result whatever it is. When a job turns out to be the wrong pick, we say so and choose another.
Who is it for?
Teams of roughly five to fifty people in New Zealand and Australia that have bought Claude or ChatGPT, or are about to, and want it used on real work rather than tried once. It is not for one person learning on their own: a free self-paced course is the better place to start.
Start with two workflows
Tell us which jobs eat your team's week. We will say which two we would start with and how we would time them.
Book a call