Human and AI Roles in the Workflow

Outcomes

• A simple view of how people and AI should work together in one real workflow.
• Agreement on who to ask when AI output is wrong or unclear
• A few plain rules you can adopt without a transformation programme

Agenda

1. 01. Framing, why roles matter more than tools

   25 min
   • When “faster” breaks handoffs, quality, or your ability to explain what happened
   • Examples: drafting, analysis, and customer comms with different risk levels
   • Pick one workflow from the room to work through together

2. 02. Mapping, who does what today

   50 min
   • Walk the workflow step by step: inputs, decisions, outputs
   • Tag steps: human-only, AI-assisted, candidate for automation
   • Surface grey zones: judgement, sign-off, exceptions, client-facing narrative

3. 03. Redesign, shifting responsibilities

   50 min
   • What moves to AI support vs. what stays with named people
   • How review and spot-checks change when volume goes up
   • What to keep on record so clients (and any regulator) can follow the thread

4. 04. Light-touch guardrails

   30 min
   • Simple standards: prompts, approved tools, and what never goes in
   • When to stop and escalate instead of guessing
   • How to tell the rest of the team without burying them in process

5. 05. Wrap-up, commitments

   15 min
   • Capture decisions and open risks
   • Suggested owners and next working session (if needed)

Format & preparation

• • Bring a real workflow (or an anonymised version). Best when it’s your actual day job, not a textbook example.
• • Works well from a handful of people up to ~20; larger groups use breakouts.