Buying AI Seats Is Not the Same as Adopting AI

Most businesses that "adopted AI" have bought it. Somebody approved a Microsoft Copilot or ChatGPT Team subscription, an email went out, and the invoice now arrives every month. Whether the work changed is a separate question, and it is the one that decides whether the money was well spent.

The gap between a licence and a habit

A seat is a permission. A habit is a person reaching for the tool without being asked because it makes their job easier. Between the two sits a lot of unglamorous work that nobody budgets for.

Think about what happens when a business buys accounting software. Nobody expects the bookkeeper to figure out bank feeds, GST codes and BAS reporting by clicking around. There is setup, there is a first month of someone checking the numbers, and there is a person who knows how it is meant to work. AI tools get none of that treatment. They arrive with a login and a hopeful message from the boss.

The result is a familiar split. One or two enthusiasts use the tool daily and get impressive results. A middle group tries it, gets a mediocre answer on a vague question, and concludes it is overrated. The rest never open it. Six months later the owner sees the invoice and wonders what happened.

Why people stall

The reasons are mundane, which is why they get missed.

• They do not know what to ask. A blank chat box is intimidating. "Use it for anything" is not guidance. People need a short list of jobs that suit their role.
• The first result was poor. Someone pastes in a half-formed request, gets a generic answer and files the tool under "not useful". The tool was fine and the request was thin.
• They are unsure what is allowed. Can I paste a client's name into this? What about a contract? Without a clear answer, careful people avoid the tool and careless people use it anyway.
• It adds a step. If using AI means leaving the system where the work lives, copying text across and pasting results back, it loses to the old way on a busy day.
• Nobody around them uses it. Habits spread through teams the way they always have, by watching a colleague do something faster.

None of these is solved by a second licence or a better model. They are solved by someone sitting with the team and the actual work.

What adoption looks like in practice

Adoption starts from the work. Ask each person which three tasks they dislike or repeat most. Sort the answers and you will see clusters: summarising long email threads, drafting first versions of quotes or proposals, tidying meeting notes, pulling facts out of documents, answering the same customer questions.

Then pick a small number of those tasks and do them properly:

1. Choose the tool for the task. ChatGPT, Claude, Gemini and Copilot overlap heavily, but your existing stack matters. If the team lives in Microsoft 365, Copilot's access to your documents may beat a stronger standalone chatbot that cannot see them. Check current terms for what each plan does with your data.
2. Write a starting prompt with the person. Not for them. A prompt they helped shape is a prompt they trust.
3. Run it on real work. A live quote, a real inbox, a genuine pile of notes. Practice on toy examples does not transfer.
4. Set the boundaries in plain words. What can go in, what cannot, and who to ask when unsure. One page is enough. Our post on shadow AI covers why this matters even if you think nobody is using these tools yet.
5. Check back. Two weeks later, ask what stuck and what was dropped. Drop what failed and keep what worked.

This is slow compared with sending an announcement, and that is the point. It is also far cheaper than another year of unused seats.

How to tell whether it worked

Avoid vanity measures. A login count rises when you send a reminder email and says nothing about value. Better signals are concrete:

• Can five random team members name one task they now do differently?
• Has a recurring task visibly shrunk, such as weekly reporting or first-draft proposals?
• Are people asking each other for prompts, or sharing a tip unprompted?
• Is the quality of output checked by a human before it reaches a customer?

If the answer to most of those is yes, you have adoption. If it is no, you have a subscription.

Also be honest about the seats that should go. If a role has no good use for the tool, cancel it. A smaller set of seats that people use beats a full set that nobody does.

Where to go from here

If you recognise this picture, the fix is organised support around your own people and tools, not a bigger budget. 21Labs, a Sydney agency serving businesses across Australia, runs exactly this kind of work under AI adoption: finding out who uses what, choosing the jobs worth handing to the tools you already pay for, and building the habit with your team. Discovery is free and we reply to every enquiry within 24 hours.