Most business owners we speak to have the same conversation with themselves about AI. A few people on the team are already using it and getting real value. The question is what to do for everyone else.
The obvious route is a licence per person: Copilot, ChatGPT Team, Claude for Work. For a team of five, that’s an easy decision. For a team of fifty or a hundred and fifty, it becomes a real line in the budget, and much of it goes to people who open the tool twice a month.
There is another way to think about it.
Seats pay for access. You want results.
A per-seat licence pays for each person to have a chat window. What most businesses actually want is narrower and more valuable:
- The salesperson gets a reminder that a key customer has gone quiet, with the last conversation and a follow-up already drafted.
- Operations finds out an order will be late before the customer chases it.
- Finance gets the month-end figures pulled together for review instead of copied by hand.
- The owner sees which commitments from Monday’s meeting have slipped.
None of that needs everyone to have their own AI subscription. It needs AI that can see your business and bring the right thing to the right person.
The real gap is connection, not chat
A general assistant knows a lot about the world and very little about your company. Microsoft Copilot sees what lives in Microsoft 365. ChatGPT and Claude see what you paste into them. Your business also lives in your order system, your accounting package, your price sheets, your shared drives and your team’s chat groups.
Until AI can reach those systems safely, every answer depends on someone copying information into a chat window. That’s the chore we were trying to remove.
So the first job isn’t choosing an assistant. It’s connecting your systems, with the same permissions your people already have, so any assistant can use them.
Two ways in, one set of connections
Once your systems are connected, there are two sensible ways for your team to use them:
- Bring your own assistant. If some of your team already pay for Claude, Copilot or ChatGPT, they plug into the same secure connections. Their assistant can now answer questions about your actual business, and it only sees what that person is allowed to see.
- A shared team workspace. Everyone else uses one workspace that runs in your environment, on the AI model you choose. You pay for what the team uses, not a licence per person, so the people who use it twice a month cost almost nothing.
Both sit on the same connections, the same memory of your business and the same record of every action. You can start with one and add the other.
What to check before you commit
Whichever route you take, ask these questions first:
- Where does our data go? It should stay in your environment or a cloud you choose, and it should never be used to train public models.
- Who can see what? AI should follow the permissions you already have. A junior team member should not be able to ask the assistant about payroll.
- Who decides? For anything that leaves the building, such as a customer email or a price, an agent should draft and a person should approve.
- What can we see afterwards? Every action should be logged: what was read, what was drafted, what was sent.
- What happens if we leave? You should own your data and your setup, and be able to take both with you.
A practical starting point
Don’t start with a company-wide rollout. Start with one team and one recurring chore that crosses two or three systems: a weekly report, customer follow-ups, a daily price update. Connect those systems, give that team the workspace, and measure the hours it returns. Then expand to the next team on evidence, not enthusiasm.
This is how we set up Nine Kites Cue: your systems connected once, usable from the assistant your people already have or from a shared workspace, with your data and your setup staying yours. If you’re weighing up AI licences for your team, book a 30-minute walkthrough and we’ll show you both routes working.