The usual story goes like this. The owner reads about Claude, buys a handful of seats, sends a link round and says have a go. Two people use it constantly, six try it once, and by the third month it is a line on the card statement that nobody quite remembers agreeing to. The tool did not fail. The rollout did.
Set up properly, Claude becomes the place a twenty-person team drafts, summarises, checks and thinks. That takes a fortnight of deliberate work, and it breaks down into four parts.
The right plan, for the right reason
For most businesses under fifty people the Team plan is the right choice. It gives everyone their own workspace, shared projects, and the assurance that business data is not used to train models. Enterprise matters when your IT policy demands single sign-on or audit logs, which for a twenty-person firm is usually a question of whether you work in a regulated sector rather than a question of size. We pick the plan with the owner and the person responsible for IT, and we write down why, so the decision survives the next renewal.
Guardrails before anyone logs in
Data controls are set once, at the start, or not at all. Which documents Claude may see. Which it may not. Who can create shared projects and who can only use them. What people are told about pasting client information, and where the line is for a firm that handles personal data under GDPR or client money under FCA rules. None of this is complicated, but it has to be decided by the business rather than by whichever employee happens to be first in.
Connected to the work
Claude on its own is a very good assistant. Claude connected to Google Workspace or Microsoft 365, to the CRM and the file store, is a colleague who has read the brief. The connectors exist and are officially supported; the work is choosing which ones to switch on, testing that permissions behave as expected, and building the two or three shared projects that hold the context the team actually uses: the tone-of-voice guide, the standard terms, the product list, last quarter's numbers.
This is where most self-serve rollouts stop short, and it is the difference between a tool people try and a tool people rely on.
Training, role by role
A single all-hands demo does not work. What works is two half-day sessions, split by what people do. Finance learns to reconcile, summarise and draft the awkward email. Sales learns to prepare for a call from the CRM record and write the follow-up in the firm's voice. Admin learns the ten prompts that remove an hour from the day. Each session ends with every person having built one thing they will use on Monday.
What it costs and what it returns
A fortnight of work, and the seat licences you were going to buy anyway. In return, a team where usage goes up over the first quarter instead of down, and an owner who can say exactly what the tool is for. The seats are the cheap part. The setup is the bit that makes them worth paying for.