Introducing a tool to other people

The two ways this usually goes wrong

The mandate. Leadership buys licences for everybody and announces that the organisation is now using AI. Adoption is measured by logins, which are high in week one and near zero by week six. Nobody says so, because saying so is awkward.

The enthusiast. One person uses it constantly, tells everyone how good it is, and produces exactly the resistance you would expect. The people who were unconvinced are now also annoyed.

Both fail for the same reason: they start from the tool rather than from a task somebody already finds tedious.

The approach that works

Find the two roles where it is transformative. In almost every organisation there are one or two jobs where these tools change the day: usually somebody who lives in a mailbox, writes repetitive documents, works across languages, or reads long documents for a living. Start there.

Solve one real problem, visibly. Not a demonstration. Take a task that a colleague complains about, do it with them, and let them keep the result. One person saying “this saved me an afternoon on the quarterly returns” moves more than any presentation.

Share the instruction, not the enthusiasm. The valuable artefact is the saved prompt with the context and the example in it, and it is transferable in a way that skill is not. A shared folder of five working prompts is the single most useful thing an early adopter can leave behind.

Let it spread by request. The second and third people should come to you.

Taking the sceptic seriously

The colleague who resists is often right about something specific, and it is worth finding out what.

• ”It makes things up.” Correct, and the answer is which tasks are checkable, not reassurance.
• ”I don’t trust it with client data.” Correct, and the answer is a rule about what is sent and to which tool.
• ”It writes badly.” Often correct, and the answer is that a first draft is not a final one.
• ”It will be used to cut jobs.” This is not a technical objection and it should not be answered with a technical response. If you do not know the answer, say so.

Somebody who has thought hard enough to object usually makes better use of the tool eventually than somebody who accepted it uncritically, because they know where it fails.

Training that is worth running

If you do run a session, make it an hour, hands on, using the team’s own work. Nobody learns this from slides. The format that works: everyone brings one real task they find tedious, and the hour is spent doing those tasks rather than demonstrating features.

Three things are worth covering explicitly because they are what people get wrong: giving enough context, supplying an example of what good looks like, and knowing which answers must be checked. Everything else can be discovered. Those three are the difference between a colleague who concludes the tool is useless and one who gets something out of it in week one.

What to write down for other people

Three short documents cover it:

1. What we use, for what. Named tools and named tasks. Two paragraphs.
2. What never goes in. Client identifiers, personal data, credentials, anything under a confidentiality clause. Five lines.
3. A folder of working prompts, each with a one-line note saying what it is for.

The third is the one people actually use. The first two are what stop the practice going wrong.

The colleague who is already using it quietly

In most organisations, somebody is. They are using a personal account on a personal device because there was no approved route, and they are not going to mention it while the position is unclear.

That person is the most useful ally available and the largest unmanaged risk, and both facts have the same remedy: make it safe to say so. An amnesty (tell us what you are using and we will find you a proper route) surfaces the real picture in a day, which no policy exercise will. It also usually reveals the two roles where the tools are transformative, because those people found out first.

The measurement conversation

Somebody will ask whether it is working. The honest answer takes two weeks of the line-a-day record from earlier in this course, kept by three or four people, and it produces a distribution rather than a number.

That distribution is the useful finding, and it is nearly always the same shape: substantial gains for a few roles, modest for most, negative for one or two where the checking exceeds the saving. An organisation that knows which is which spends its money well. One that has an average does not know anything.