"We gave the whole team AI" usually means somebody bought assistant licences and waited for something to happen. An augmented role is a different thing, and you recognise it by how one person's week changes.
Before and after a week
Before: the advisor spends Monday morning preparing three meetings, Tuesday writing the summaries of those meetings, and Wednesday chasing data for a quarterly report.
After: on Monday they review three briefing packs that are already assembled and adjust the angle of each conversation; on Tuesday they approve summaries drafted from their notes; on Wednesday the report draft is waiting and they work on the argument, not the figures.
The number of hours doesn't change. What they are spent on does.
The four pieces
- The agent: it runs a complete task, it doesn't answer loose questions.
- The access: to the real data and tools, or the agent works blind.
- The boundary: what runs alone, what needs approval, what gets logged.
- The training: the person has to know when to distrust the output.
Miss one of the four and you get a pleasant experiment that is abandoned in month two.
Why a role and not a department
A role has an owner and an observable routine. A department has meetings. When the scope is a role, the conversation stops being about AI strategy and becomes about what the agent did on Tuesday morning — which is where the learning is.