Workflow design · Illustrative case study
Designing for the conversation, not just the seating chart.
A dinner has a small number of seats and a surprisingly large number of decisions. The useful work is making those decisions explicit before asking AI to help.
Explore Nine Tables, One System(opens in a new tab)Context
Co-leading the SF chapter of Women + AI has given me a practical setting for using AI in community operations. This case study explains an approach reflected in the shared chapter workflows I use. It does not claim sole authorship of those workflows or reproduce the chapter's private materials.
The problem
A guest list alone does not tell an organizer who is confirmed, who can facilitate, which constraints matter most, or whether a last-minute swap has left someone out. A plausible-looking plan can still be wrong.
The design decision
Separate eligibility and hard constraints from preferences. Ask the organizer to resolve missing facts. Only then propose a plan, and finish by reconciling every eligible guest against the output.
- Confirm the source roster, attendance status, and designated facilitators.
- Count eligible guests and check table capacity before placing anyone.
- Apply hard constraints, then balance functions and conversation interests.
- Show tradeoffs and unresolved inputs for organizer review.
- After edits, check every guest appears exactly once and every table still meets the rules.
Explore the example
Fictional example: 12 confirmed guests, two tables of six, and two confirmed facilitators. The example checks only capacity, coverage, and facilitator count.
Table 1
- Guest A · Facilitator
- Guest B · Product
- Guest C · Engineering
- Guest D · Operations
- Guest E · Research
- Guest F · Design
Table 2
- Guest G · Facilitator
- Guest H · Engineering
- Guest I · Product
- Guest J · Design
- Guest K · Operations
- Guest L · Research
Coverage check: all 12 guests appear once, each table has six seats and one designated facilitator. Company and preference constraints are outside this simplified example.
What the work demonstrates
The reviewed workflow specifications include explicit counting, facilitator confirmation, exception handling, and completeness checks. That establishes the design of the process; it does not establish a measured improvement in attendance, time saved, or conversation quality.
Scope and limits
This page's worked example is fictional. It is not a live roster integration or an automatic seating optimizer. Shared skill files and real event data are not distributed here.