Most workforce plans start with jobs. They take today’s roles, apply an assumption about what AI or automation will do to them, and arrive at a number. The result is usually a headcount figure nobody quite believes and a workforce that’s anxious about what it means.

There’s a better starting point: the work itself. When you redesign work at the level where value is created – tasks, decisions and handoffs – the implications for roles, skills and structure follow from the design rather than being imposed on it.

This guide sets out how to approach work redesign for a single workflow, in weeks rather than months.

Why start with work, not jobs?

Jobs are bundles of tasks that made sense when they were designed. AI doesn’t replace jobs neatly; it changes some tasks within many jobs, makes others possible for the first time, and creates new ones – reviewing output, handling exceptions, owning data quality.

Start with jobs and you miss most of that. You end up asking “how many of these roles do we still need?” instead of “how should this work be done now?” The first question produces anxiety and resistance. The second produces a design people can understand, test and run.

“The unit of change is work, not workforce. Workforce implications follow from the design.”

The 8-step work redesign method

Step 1: Choose one workflow that matters

Resist the urge to redesign everything at once. Pick one workflow where improvement would be visible to customers, finances or risk – claims handling, supplier onboarding, month-end close, customer complaints, a planning cycle. It should be important enough to matter and bounded enough to finish.

Write down what better would look like in business terms before you start: faster, cheaper, fewer errors, better decisions, a better experience for customers or staff. Baseline those measures now.

Step 2: Map the work with the people who do it

Not with their managers, and not from the process documentation. Sit with the people who do the work and map what actually happens: every task, every decision, every handoff, every workaround and every place where someone waits for something.

This step does 2 jobs. It produces a far more accurate picture than any document will. And it tells the people closest to the work that the redesign is being done with them rather than to them – which matters a great deal for adoption later.

Step 3: Sort the work

Take each task and decision and place it in one of 3 categories:

• Human judgement – work that needs context, discretion, relationships, ethical weight or accountability. Negotiating, deciding on exceptions, handling a distressed customer, challenging an assumption.
• AI or automated execution – work that is rules-based, high-volume or synthesis-heavy. Gathering information, drafting, checking against standards, summarising, routing.
• Shared, with review – work where AI does the first pass and a person reviews it. Set the review threshold by risk: what level of error is acceptable, and what must always be seen by a human?

Be honest about the middle category. It’s where most of the redesign lives, and most of the risk.

Step 4: Redesign the flow, not just the tasks

Now redraw the workflow. Remove steps that only existed to move information between people. Bring decisions closer to the point where the information is. Make human review deliberate rather than habitual.

Pay most attention to the handoffs. They’re where work gets lost, delayed or duplicated – between people, between teams, and now between people and AI. Every handoff should have a clear owner and an explicit acceptance.

Step 5: Let the roles follow from the work

Only now look at roles. Group the redesigned tasks and decisions into coherent jobs. Some roles will change shape significantly. Some new ones may appear – people who oversee AI output, own data quality, or handle the exceptions automation surfaces.

This is where workforce implications become clear, and where they become defensible. You can explain every change by pointing to the work, rather than to an assumption.

Step 6: Design the transition from the start

People need to know what is changing, why, and what it means for them – early. Plan capability building, changes to management routines and incentives alongside the design, not after it.

Ask whether the organisation can absorb this change alongside everything else it’s carrying. If it can’t, sequence it. Change capacity is finite, and overloaded teams adopt nothing well.

Step 7: Test it in the real world

Run the redesigned workflow with a real team for a few weeks before scaling it. Measure against the baseline from step 1. Expect to adjust: review thresholds will be wrong, a handoff will be clumsy, a task you classed as automatable will turn out to need judgement.

Treat these as findings, not failures. They’re exactly what a test is for.

Step 8: Leave the capability behind

The final test of a work redesign is whether the organisation can do the next one without outside help. Hand over the method, the maps and the decision logic, and make sure someone internally owns them.

Common mistakes to avoid

1. Starting with a headcount target. It shapes every subsequent decision and makes the people you need most defensive.

2. Redesigning from the process manual. The documented process and the real one are rarely the same.

3. Automating a broken workflow. You get the same problems, faster.

4. Reviewing everything by default. Blanket human review wipes out the gains and dulls attention where review really matters.

5. Treating training as the transition plan. Training is one part. Incentives, management routines and decision rights matter as much.

A note on headcount

Work redesign will sometimes change how many people a workflow needs. That conversation goes far better when it follows from a design people helped build and can see the logic of. Start with headcount and you get resistance. Start with the work and you get something people can actually run.

Common questions

What is work redesign?

Work redesign is the process of reshaping how work is done – tasks, decisions, handoffs and roles – to improve outcomes, often around new technology such as AI. It starts with the work itself rather than with existing jobs or org charts.

How does AI change job design?

AI changes tasks within jobs rather than replacing jobs wholesale. It takes on high-volume, rules-based and synthesis work, shifts people towards judgement and exceptions, and creates new tasks such as reviewing AI output and managing data quality.

How long does a work redesign take?

For a single, bounded workflow, a redesign and real-world test can take a matter of weeks. Larger, organisation-wide changes are best approached as a sequence of workflow redesigns rather than one programme.

Should employees be involved in work redesign?

Yes. The people doing the work know how it really runs, and involving them produces a more accurate design and much stronger adoption.

What this looks like in practice

When a national health insurer replaced a decades-old CRM, we embedded a ways-of-working practitioner who baselined how teams actually worked and ran a backlog of small experiments in roles, planning and prioritisation. The practices were later adopted by other teams on subsequent programmes – the capability outlived the engagement. Read the case study →