Rolling out an enterprise AI assistant: lessons for large companies

Many large companies have bought enterprise AI assistants, such as those built into office software, for some or all of their staff. A common pattern follows: strong interest in the first weeks, then usage concentrates in a small group.

Before launch

• Permissions: review who can access which files and sites. An assistant that searches company documents will surface anything a user technically has access to, including over-shared files.
• Policy: publish what the assistant may be used for and what data rules apply.
• Use cases: pick three to five concrete tasks per department, such as summarising meetings, drafting emails, or preparing reports.

During rollout

• Launch in waves, starting with teams that have clear document-heavy work.
• Provide role-specific training and a library of example prompts.
• Appoint champions in each team and hold regular clinics.

After launch

• Track active usage by team and follow up where it is low.
• Collect time-saving examples and share them internally.
• Move from individual productivity to automated workflows, where the larger savings are.

The missing role

Most stalled rollouts lack someone accountable for adoption and value. An AI implementation lead or enablement lead for the first six to twelve months is usually the difference between a licence cost and a productivity gain.