Redesigning work
The biggest gains come from redesigning how work flows — with the people who do it. This module shows how.
Learning objectives
- Explain why workflow redesign beats tool roll-outs
- Analyse a role at task level
- Co-design new workflows with human checkpoints
- Plan for freed time and guard against deskilling
4.1 From tool to workflow
Handing people an AI assistant and hoping for productivity gains gives modest results. Bigger gains come from redesigning workflows: deciding which steps AI does, which people do, and how they hand over. Redesign is also where most of the change effort — and the risk of getting it wrong — lies.
4.2 Break jobs into tasks
Jobs are bundles of tasks. AI affects tasks, not whole jobs. For a role, list its main tasks and assess each:
| Task type | Likely approach |
|---|---|
| Routine, rules-based, low-stakes | Automate (with monitoring) |
| Drafting, searching, summarising | AI drafts, person reviews and finalises |
| Judgement, relationships, accountability | Person leads; AI provides information |
| Creative, strategic, novel | Person leads; AI as a thinking partner |
4.3 Redesign with the people who do the work
- Map the current process with the team, including workarounds
- Identify pain points and where AI could help (Module 2 of AI Business Strategy)
- Design the new flow: AI steps, human checkpoints, handovers and exceptions
- Pilot it, measure, and adjust before rolling out
- Update procedures, training, quality checks and performance goals to match
People who help design a change are far more likely to make it work.
4.4 What happens to freed-up time?
Decide deliberately. Options include more time with customers, tackling backlogs, improving quality, developing skills, or absorbing growth without extra hiring. If you don't decide, the time tends to disappear into low-value activity — and the business case (see AI Business Strategy, Module 3) won't be realised.
4.5 Keep humans skilled and accountable
When AI does the routine work, people can lose the skills they need to check it or to step in when it fails ("deskilling"), and may come to accept its output without thinking ("automation bias"). Counter this with regular practice without AI, rotating review duties, clear accountability for outputs, and training in validating AI output.
Choose one process your team runs. List its steps, mark each as automate / AI drafts + human reviews / human leads, and sketch the new flow with human checkpoints. Note what you'll do with the freed time.
See a worked example
Example — monthly supplier report: Gather data (automate from systems) → AI drafts commentary on variances → analyst checks figures and edits commentary → manager reviews flagged risks → send. Freed time (~1 day/month) goes to deeper supplier-risk analysis.
Key takeaways
- Redesign workflows, don't just add tools.
- AI changes tasks, not whole jobs.
- Design with the people who do the work.
- Decide what freed time is for.
- Keep people skilled and accountable; watch for automation bias.