AI at Work

Organisation, then department, then the people who carry it.

The work runs from the whole organisation down to each department, then to the few people who will carry it once we have gone. Evidence comes first at every level.

The sequence

We start with your own evidence

Before anyone stands up in a room: who holds which licence, which tools are on the approved list, where documents live, and what people are already doing with the assistants, sanctioned or not.

The licence position is settled before the first session, so nothing is shown that the room cannot use on Monday.

A small cross-functional group goes first

Operations, finance, IT, and whoever owns data classification. What they struggle with sets the organisation-wide decisions on policy, education and licences.

Each department names its jobs

Delivered by department, with the manager in the room, not in mixed cohorts. Each department picks the three jobs it will change next, and each job is followed from the file it starts in to the place it lands.

For every job: which parts go to the assistant, which get automated, which are calculated outside it, and who checks.

A working method is left behind

Templates saved, checking rules written, prompts built against real documents, and named internal leads who go deeper than everyone else. If nothing is different by the end of the session, the session failed.

If a job needs a build rather than a method, that is said plainly, before seats are bought or anything is commissioned.

Sessions, not a catalogue.

We work across Copilot, ChatGPT and Claude, inside whatever your organisation has approved.

Leadership session

Half a day. Where the value is, which jobs first, what the licence position should be.

Departmental working session

Two to three hours, hands-on, on the department’s own files, manager present.

Modular series

A recurring working session for a pilot group, each one rebuilt from the last one’s feedback.

Deep dive

A narrow group at the end, for the people who will carry it afterwards.

What we do not do

  • Sell or resell licences, or take referral fees from anyone whose tools we assess.
  • Certify people against a vendor curriculum.
  • Demonstrate features the room is not licensed for.
  • Name tools outside your approved list in a session.
  • Report adoption as a login count, or publish a number without saying where the effort went.

Questions about fit: hello@acuityai.co. Or try it on three jobs first.