In 30 seconds

This is a reading path, not a single lesson. It takes you from the messy reality of a real workflow to a tested pilot and an honest decision about whether to deploy.

Expect to spend the time across several sessions, one stage at a time.

Why a path helps

The individual articles each answer one question. This path connects them in the order the work usually happens, so you always know where you are and what comes next.

You can stop at any point. A good assessment that concludes "AI is not the right tool here" is a successful outcome, not a failure.

The path

1. See the whole shape

Read The Workflow Ladder. It gives you the eight stages and a place to locate your own project.

2. Understand the real work

Read How to interview someone about their work. Walk through one recent, real case with the person who does it.

3. Make the workflow visible

Read How to map a workflow. Turn the interview into a simple map of steps, decisions, handoffs and exceptions.

4. Find where the truth lives

Learn what a source of truth is and confirm which system or document is actually current.

5. Make it digital enough

Read How to turn a manual process into a digital workflow. Fix the foundations before adding intelligence.

6. Decide where AI helps

Read How to decide whether a workflow needs AI and Automation vs AI vs agent.

7. Design the behavior and the boundaries

Learn read access vs write access and suggest vs draft vs execute. Decide what the system may read, decide and change.

8. Build the smallest useful pilot

Read How to build the smallest useful pilot and How to run a shadow-mode pilot.

9. Evaluate real cases

Read How to create a test set from real work. Measure quality against actual work, not a demo.

10. Decide whether to deploy

Proceed, redesign or stop. If you deploy, read How to assign ownership after launch.

What this does not solve

A path cannot make a bad use case good. If the underlying workflow is not understood, no later stage will rescue it. That is why the early stages come first.

Good enough to move on when

  • You have chosen one real workflow to follow through the path.
  • You know which stage that workflow is currently in.
  • You are ready to start with understanding, not with a model.