The ten workshops
Workshop 1 is run once per use case, with the SMEs for that use case. Northbridge Consulting runs it three times: proposals (Dana Ortiz), staffing (Sam Whitfield) and contracts (Helen Park). Plan the calendar with that in mind.
Dependency order
Each workshop depends on decisions made in earlier ones. A data mapping written before the ontology exists has to be rebuilt. Matching rules cannot be written until you know which entity types exist and which keys each source carries. Read the diagram left to right:- Discover (WS1, WS2) comes first. Nothing else can be designed until you know the questions and where their answers live.
- Model (WS3 to WS7) turns those findings into configuration. WS4 and WS6 can run in parallel once the ontology draft exists.
- Pilot ingestion loads real data. WS8 and WS9 need a populated graph to be useful.
- WS10 comes last, because assistants and agent flows are only as good as the graph beneath them.
Which workshops are optional
- WS9 Access Control is optional. Run it only when the firm has ethical walls, confidential contracts or other record-level visibility rules. See access control.
- WS8 Enrichment Rules can be very short if the pilot shows the documents already carry the attributes you need. Do not skip the conversation; confirm there is nothing to infer.
- Agent flows in WS10 are optional for a first launch. Most firms start with a chat assistant and add agent flows once the golden questions pass.
Cadence
Map the workshops onto the roadmap:
Leave at least one working day between modeling workshops. The FFE needs that time to enter the configuration, and the super user needs it to review what was entered before the next session builds on it.
Expect a second pass
Modeling workshops iterate. The first pass of WS3, WS5 and WS7 is based on what people say the data looks like. Pilot ingestion shows what it actually looks like: resumes that name people differently from HR, SOWs without project codes, a document family nobody mentioned. Book a 90-minute “model revisit” session after the first pilot ingestion, and tell the client at kickoff that it is planned, not a sign of failure.Roles in the room
The super user and the SME for the use case together own each decision. The FFE should propose, not decide. If the FFE makes business decisions, the client cannot maintain the configuration after handover. See Roles and Permissions for the Experio access each person needs.
Preparation rhythm
Every workshop follows the same rhythm.1
Five working days before: book and brief
The PM sends the invite with the purpose, agenda and attendee list. Name the decisions the session must make, not just the topic.
2
Three working days before: send pre-reads and homework
The FFE sends the pre-read: the previous workshop’s outputs, a draft of what this session will produce, and specific homework (for example, “bring five real questions you answered last week” or “export 20 sample rows from Deltek”).
3
One working day before: prepare the draft
The FFE drafts the likely answer from the homework: a draft ontology, a draft mapping, a draft taxonomy. Reacting to a draft is much faster than designing on a blank whiteboard.
4
Day of: run the session
Follow the workshop page’s agenda. Close with a read-back of every decision.
5
Within one working day: send the follow-up
The PM sends the decision log, parking lot and action owners. The super user replies to confirm sign-off.
Good pre-reads
- Short: two pages or less, plus attachments.
- Concrete: real file names, real column headers, real questions.
- Marked with what the reader should do: “Check that these 12 entity types cover your questions” beats “please review”.
Capturing decisions
Keep two living documents for the whole implementation. The PM owns both; the templates are on the Templates page.Decision log
One row per decision. Record the decision, not the discussion.
The last column matters. A decision is not done until it is configured and the super user has seen it in Experio.
Parking lot
Anything that is important but not for this session goes here, with an owner and the workshop where it will be resolved. Review the parking lot at the start of every session. Items that stay in the lot for three sessions either get a decision or get dropped.Remote, in person, or hybrid
For remote sessions, keep to two hours with a break, share one screen only, and have the PM post each decision in the meeting chat as it is made. Ask people to keep cameras on for decision points.
Handling disagreement
Disagreement is useful. It usually means two teams use the same word for different things, which is exactly what the ontology must resolve.- Name the disagreement. “Sales calls it an account; Finance calls it a client. Are they the same thing?”
- Go to an example. Ask for a real record. Arguments about definitions end quickly when you look at the Lakeshore Health row in both systems.
- Go back to the questions. Which option lets the golden questions be answered? The questions are the tie-breaker.
- Decide reversibly. Most modeling choices can be changed after the pilot. Say so, pick one, and log it.
- Escalate if needed. If the super user and SME still cannot agree, park it with the sponsor (at Northbridge, Priya Shah) as the owner and a date.
Closing each workshop
Save the last 15 minutes of every session for closing. Do not let it get squeezed.- The PM reads back every decision in the log.
- The decision owner confirms each one aloud, or it goes to the parking lot.
- Review the workshop’s exit criteria (on each workshop page) and tick what is done.
- Agree the actions, owners and due dates, including client homework for the next session.
- The super user confirms sign-off by replying to the follow-up email. That reply is the record.