Example scenario, not a client project. A logistics company wants AI to answer shipment-status queries. The map shows that 60% of the team's time goes not into answering, but into looking up data in three systems and chasing the warehouse. So the first step isn't a chatbot — it's bringing the status into one place, with AI preparing a draft reply that an employee approves.
Is this for you?
- The procedure says one thing, people do another, and no one knows exactly what.
- The same case passes through several teams and systems, and gets lost along the way.
- You want to introduce AI or an agent, but you don't know where in the process.
- The team works in spreadsheets, email and chat threads, because the tool doesn't cover everything.
- A new hire needs months to understand how it really works.
What I do
- Operational processes — request handling, documents, approvals, billing, work planning.
- Product journeys — how a user moves through the product to get a task done, including what they do outside it.
- Handoffs — where work passes between people, teams and systems, and what gets lost when it does.
The AI layer
- I tag every step: repetitive or requiring judgment, what data it uses, what an error costs.
- On that basis I show where AI makes sense (e.g. classification, summarizing, drafting), where it can at most make suggestions, and where the decision has to stay with a person.
- I check where people already use personal AI tools today — often the best clue to where the real need is (MIT NANDA: 90% of employees use them on their own).
- I estimate the cost of leaving the process as it is: time, rework, delays.
What you get
- A current-state workflow map — steps, roles, tools, handoffs, workarounds.
- A waste map — where work waits, comes back for rework or is duplicated.
- An AI readiness rating for every step — suitable / with a person / not suitable.
- A short list of places to change, with the reasoning.
- A process glossary — names the team starts to understand the same way.
How it runs
- Kick-off conversation — we agree which process matters, and whose point of view matters most.
- Materials and data — procedures, screenshots, sample cases, data from your systems.
- Interviews and observation — a few sessions with the people who do the work (a short version of Research).
- Working map — a first version the participants correct.
- Review session — together we confirm the map and the places to change.
Roughly 2–3 weeks per process [to be confirmed].
How we'll know it worked
- The team and leadership agree that the map shows the real work.
- It's clear which 2–3 places we change first, and why.
- AI ideas are tied to specific steps, not to "the whole process".
What this doesn't cover
I don't implement tools or automation, and I don't write ISO procedures. The map is the basis for decisions and for design.
Visualizations
The procedure vs. reality
The real workflow includes the steps nobody wrote down.
AI readiness map (sample artifact)
Sample artifact: each step rated for where AI fits.