“We need to do something with AI” can open a conversation, but it leaves a consultant with a lot to untangle. Your first job is to help the client describe a piece of work worth improving and the decision they need to make.
Discovery that reaches a decision.
1
Start
A recent case and desired outcome
2
Understand
People, evidence and constraints
3
Choose
A next step and accountable owner
Start with a recent piece of work
Ask: “Could you walk me through the last time this became a problem?” A recent order, case or request gives everyone something concrete to examine. Ask what started it, where it went, who became involved and what counted as finished.
Then ask what the client wants to improve. Faster first drafts, fewer returned orders and shorter customer wait times describe different outcomes. Agree which one matters before comparing tools.
Six questions that make the opportunity clearer
1. What should be better for the customer, employee or business? Ask how the sponsor would recognize improvement, and what must not get worse.
2. Where does the work slow down, return or leave the normal process? Ask for examples from both ordinary cases and exceptions.
3. Whose knowledge makes it work today? Look beyond the person officially assigned. A colleague may routinely interpret incomplete information or rescue difficult cases.
4. What has already been tried? Separate tool performance from access problems, unclear instructions, missing information and time to practice.
5. What evidence can we examine? Establish what records exist, what they omit and who can authorize their use.
6. Who can approve a change and support a test? A promising idea needs a decision owner, affected participants and capacity to follow through.
A manufacturing example
Imagine a sponsor wants AI to write quotations faster. Sales demonstrates quick draft generation. Production then explains that quotes are returned because delivery dates assume capacity that nobody confirmed. An experienced planner has been resolving those conflicts by phone.
The problem is now more specific: how should a quote obtain a reliable capacity decision before it reaches the customer? AI drafting may still help, but the discovery needs to examine the planning handoff, exception rules and the information people need to decide. This example is illustrative, not a reported Gobekli deployment.
Separate what you heard from what you know
Capture observations, interpretations and open questions separately. “Three sampled quotes needed a planner’s correction” is an observation. “Capacity assumptions cause most delays” is a hypothesis requiring more evidence.
Invite participants to correct the picture. Explain what will be shared and how it will help their work. Avoid turning a discovery conversation into an unexplained assessment of individual performance. Your goal is a usable account of the workflow, including disagreements that need resolution.
Close with a decision, not a long wish list
Summarize the workflow, intended outcome, suspected breakdown and missing evidence. Recommend a focused working session only if it can help the client resolve those questions. If a technical test is needed first, say what it would establish and who should run it.
Gobekli can help you shape that discovery into a scoped engagement using your method. Bring a recent example, a sponsor and the question the team needs to answer. Together, we can decide what support and outputs would make the work worthwhile.
Try this with one client situation.
Copy these prompts into your working notes. Keep observations, assumptions and open questions separate.