Discovery workshops are where process knowledge lives. Someone explains how work gets done, and a facilitator captures it in notes, recordings, or sticky notes on a wall.
The problem comes afterward. Turning those raw notes into a clean BPMN diagram takes hours or days. Details get lost. The modeler interprets what they remember instead of what was said. By the time the diagram is done, half the nuance from the original conversation is gone.
Crismo's AI discovery feature closes that gap. You paste a transcript, and the system extracts tasks, decisions, roles, and handoffs into a structured BPMN draft. Not a finished model. A draft that captures the right structure so you can refine instead of rebuild.
The problem with manual translation
Every process modeler has been through this cycle:
- Run a 90-minute discovery session
- Take notes (or record the session)
- Go back to a desk and try to reconstruct the process from memory and notes
- Realize the notes are ambiguous on three key decision points
- Schedule follow-up calls to clarify
- Finally produce a diagram, days later
The translation step, from notes to diagram, is where most information loss happens. It is also the most tedious part of the job. The discovery conversation was rich and specific. The notes are sparse and ambiguous. The modeler fills the gaps with assumptions.
How AI discovery works in Crismo
The workflow is three steps.
Step 1: Paste the transcript
After a discovery workshop, you have raw material: interview notes, a meeting transcript, a voice recording transcript, or even bullet points from a sticky-note session.
Open Crismo and paste the text into the AI discovery panel. The system accepts unstructured text. It does not need a specific format or template.
Step 2: Review the extracted structure
Crismo's AI analyzes the text and extracts:
- Tasks with descriptive names (not generic labels like "Step 1")
- Decision points where the text describes branching logic ("if the amount exceeds 10,000...")
- Roles and participants mentioned in the conversation
- Handoffs between people or systems
- Start and end conditions for the process
The result is a structured BPMN draft displayed on the canvas. Elements are positioned, connected with sequence flows, and assigned to lanes where the transcript mentions specific roles.
Step 3: Refine the draft
The AI-generated draft is a starting point, not a finished product. You will need to:
- Adjust naming to match your organization's terminology
- Add exception paths that the transcript only hinted at
- Split or merge tasks where the AI grouped things differently than you would
- Validate the model against what participants actually said
This refinement takes 15 to 30 minutes instead of the hours you would spend building from scratch.
When this works best
AI discovery is most useful when:
- You have a transcript or detailed notes from a real conversation (not a theoretical process description)
- The process has 10 to 30 steps with clear roles and decisions
- You want a first draft fast so you can iterate with stakeholders
- Multiple discovery sessions need to be modeled and you are short on time
It is less useful when:
- The source material is a single paragraph or a few bullet points (too little context for the AI to extract structure)
- The process is highly technical with system-level detail that requires domain expertise
- You already have a clear mental model and just need to draw it
A practical example
Imagine you ran a discovery workshop for an employee onboarding process. A participant described the flow like this:
"When we get the signed contract back from the new hire, HR creates their profile in the system. Then IT gets a ticket to set up the laptop and accounts. At the same time, facilities assigns a desk. Once everything is ready, the manager sends a welcome email with the first-week schedule. If the new hire is remote, we skip the desk assignment and ship the laptop instead."
Paste that into Crismo. The AI extracts:
- Start event: Signed contract received
- Tasks: Create HR profile, Set up laptop and accounts, Assign desk, Send welcome email, Ship laptop
- Parallel gateway: IT setup and facilities assignment happen concurrently
- Exclusive gateway: Remote vs. on-site determines desk assignment vs. laptop shipping
- Roles: HR, IT, Facilities, Manager
- End event: Welcome email sent
The draft lands on the canvas with lanes for each role, parallel paths for concurrent work, and a decision point for the remote/on-site split. You refine the naming, add an exception path for missing equipment, and the model is ready for review.
From discovery to documentation
The AI discovery feature fits into a larger workflow:
- Prepare using the process discovery workshop guide on ProcessCamp to structure your session
- Run the session and capture a transcript or detailed notes
- Paste into Crismo to get a structured BPMN draft in seconds
- Refine the draft with your domain knowledge
- Validate by sharing the model with workshop participants
- Iterate using as-is/to-be modeling to design improvements
The guide on ProcessCamp covers steps 1 and 2 in detail: how to scope the session, who to invite, which questions to ask, and how to structure the 90 minutes. Crismo picks up at step 3.
Try it yourself
Open the Crismo playground and paste any process description. No account required.
Start with a real transcript from your last discovery session. Or use the onboarding example above. See what the AI produces and how much editing time it saves compared to building from scratch.
