Redesign the Work

After understanding the work and deciding what should change, the question becomes: how should the work system be redesigned so the outcome improves — not just the task where AI enters?

Most AI programs stop at the task. They make the activity faster or cheaper without changing the surrounding system of roles, coordination, expertise, authority, and incentives. The task improves. The outcome-producing system stays the same — or degrades.

Redesign means changing how work gets done, not just which tool does it.

A work system is more than a workflow. Redesign must consider:

• Roles — who does what, and what judgment remains with humans vs. machines.
• Coordination — how people and systems hand off work, share information, and resolve conflicts.
• Expertise — how the organization builds, maintains, and applies judgment — and whether AI accelerates or hollows that out.
• Authority — who recommends, who decides, who approves, who overrides, who escalates.
• Incentives — what behaviors the system rewards, and whether those align with the outcome you actually want.

If AI changes the task but leaves these untouched, you have tool adoption, not transformation.

AI changes what activities are cheap, fast, or even possible. That creates an opportunity to rebundle work — to reorganize who does what, in what sequence, with what tools.

The temptation is to automate task by task. The opportunity is to ask: if we were designing this work system today, with current AI capabilities, how would we organize it?

Some things that look like inefficiency are actually load-bearing functions: exception handling, expertise development, coordination, quality control, institutional memory. Redesign must protect these — not by preserving legacy tasks, but by ensuring the functions have a home in the new system.

The hidden-functions analysis from Chapter 1 feeds directly into redesign. For each function identified, the redesign must specify where it will live and how it will be performed.