Beyond the List of AI Use Cases: How Leadership Teams Choose the First Workflows to Redesign

In short: Many companies collect AI use cases and work through them one by one. That creates activity, but it rarely changes how the company creates value. In the AI-first operating model, a different question comes first: which three to five workflows do we redesign from the outcome backward? They are chosen by volume, friction, information load and how easily mistakes can be fixed. Each one gets an accountable owner and a goal with a number.

Why do lists of AI use cases rarely lead to real change?

The usual question is: where can we use AI? It sounds pragmatic and almost always produces a long list. Each item improves a single step inside an existing workflow. The workflow itself, with its handoffs, approvals and waiting times, stays exactly as it was.

In the first building block of the AI-first operating model, capabilities, the work therefore starts with the outcome. That means the workflows through which a company creates value: preparing quotes, processing orders, handling a complaint. The better question is: what would this workflow look like if we designed it from scratch today, with agents in it?

Which criteria should guide the choice of the first workflows?

The World Economic Forum recommends starting with three to five workflows. Four characteristics help with the choice, and I describe them in more detail in my book:

  • High volume: the workflow runs often, so every improvement multiplies.
  • High friction: there are many handoffs, follow-up questions and waiting times between departments.
  • Heavy information load: decisions depend on pulling together knowledge from several sources.
  • Fixable mistakes: an error shows up early and can be corrected before it does serious damage.

A workflow that meets all four is a strong candidate. In many industrial companies, that’s quote preparation: high volume, lots of handoffs, and the necessary knowledge often sits in a few people’s heads.

What does the move from current process to target design look like?

In my book, I follow the fictional machinery company Overbeck through this step. The division head maps quote preparation as it runs today and counts 47 steps and eleven handoffs. After four weeks of work, the target design has 19 steps and three handoffs. An agent reads the inquiry, compares it with past cases and drafts a price proposal. The experienced design engineer only sees the cases where the agent can’t find a similar variant.

The remark that comes up along the way says a lot about the real obstacle: “AI won’t be what stops this. It’ll be that I have to give up approving quotes under 50,000 euros.” Redesigned workflows shift decision rights. That’s where the work is.

Who owns a redesigned workflow?

Most workflows that customers notice cut across several functions. That’s why every workflow needs a workflow owner, a person accountable for it end to end. The WEF names this role explicitly. In many companies, it so far exists only on paper in a process manual.

To be effective, a workflow owner needs a clear mandate, agreed by the leadership team and known to everyone involved. They also need influence over the goals of the teams involved. As long as engineering is measured only on development projects, it has little reason to answer quote requests quickly.

What should leadership teams review now?

  • One question: Which three to five workflows would we redesign if we set our list of AI use cases aside?
  • One metric: for each of these workflows, the number of handoffs and today’s cycle time, measured before the redesign.
  • One conversation: with potential workflow owners about their mandate and which approvals they would need.

Frequently asked questions

What is the difference between an AI use case and a redesigned workflow?

A use case improves a single step. A redesigned workflow reorganizes the work around the desired outcome, with fewer handoffs, new approvals and agents as a built-in part.

What does a workflow owner do?

A workflow owner is accountable for a workflow across department lines, including the agents and the knowledge they rely on. The role needs a mandate from senior management and influence over the goals of the teams involved.

What happens to the time that gets freed up?

Senior management should decide that in advance and say it openly. If freed-up time isn’t used deliberately, it disappears into daily routine, or it creates fear about jobs that slows the whole effort down.

Further reading: The AI-first operating model · AI agent autonomy · The organization after AI agents

Sources: World Economic Forum with Kearney (2026): The AI-First Operating System · Nicole Lontzek (2026): Die neue Betriebslogik, Chapter 4

Similar Posts