You get:
- tribal knowledge that leaves when employees leave
- inconsistent execution (different people, different methods)
- no training materials for new hires
- processes that can’t be improved because they aren’t documented
- wasted time figuring out how to do tasks
But process discovery is not guesswork.
It is interviewing the people who know.
- Interviews: ask about steps, decisions, exceptions
- Observation: watch the process in action
- Step extraction: break down into discrete actions
- Decision points: where choices are made
- Handoffs: where responsibility transfers
Without discovery, SOPs are incomplete.
This framework forces AI to extract processes from subject matter experts.
Assume the role of a process documentation specialist who extracts knowledge from subject matter experts. Your task is to discover and document a process. Generate: 1. PROCESS NAME AND GOAL - What process are we documenting? - What outcome does it achieve? 2. STEP-BY-STEP EXTRACTION - List of steps in order - Sub-steps where needed 3. DECISION POINTS - Where choices are made - Conditions for each path 4. ROLES AND RESPONSIBILITIES - Who performs each step - Handoffs between roles 5. EXCEPTIONS AND EDGE CASES - What can go wrong? - How to handle exceptions 6. TIME ESTIMATES - How long each step takes - Total process time 7. INTERVIEW QUESTIONS FOR SME - Questions to ask the expert INPUTS: Process Name: [INSERT] Process Owner/Role: [INSERT] Known Steps (if any, partial list): [LIST OR "UNKNOWN"] Inputs (what starts the process): [DESCRIBE] Outputs (what the process produces): [DESCRIBE] Frequency (how often process runs): [DAILY / WEEKLY / MONTHLY / AS NEEDED] RULES: - Interview questions must be open-ended (not yes/no) - Step extraction: break down into discrete actions - Decision points: if-then logic - Handoffs must be clear between roles - Exceptions: plan for what can go wrong - Time estimates: be realistic (include buffer) - Validate documented process with SME before finalizing
- Interview the person who does the process most often.
- Watch them do the process (don’t just ask).
- Ask “what could go wrong?” for each step.
- Validate the documented process back with the SME.
- Document processes before the expert leaves the company.
Process Name: New Client Onboarding
Process Owner/Role: Client Success Manager
Known Steps: Sign contract, schedule kickoff call, send welcome email, set up account, introduce to team
Inputs: Signed contract from sales team
Outputs: Fully onboarded client, active account, kickoff meeting completed
Frequency: AS NEEDED (when new clients sign)
This framework improves outcomes by forcing:
- step-by-step extraction (granularity)
- decision point identification (conditional logic)
- role and handoff clarity (accountability)
- exception planning (risk management)
- time estimation (workload planning)
Great process discovery captures knowledge before it walks out the door.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI business strategy tools, SOP frameworks, and practical strategies for leaders and operators.