2026-08-11 Executive Team

[joecastiglione]

Agenda

• Admin
  • Task Order amendment policy
  • Phase 12 Prioritization / Issue descriptions (draft template)
• Tech
  • Explicit Telecommute (WSP)
  • Task schedules
    • ActivitySim Application & Analysis Guide (CS)
    • Telecommute Frequency (MWCOG, RSG)

Notes

Explicit Telecommute Design — Implementation Update (Sijia Wang)

• Re-examined workers reporting both telecommute time and an out-of-home work tour (~4–6% of workers). Found the survey's telecommute-time question captures work anywhere (e.g., coffee shops, co-working spaces), not just at home.
• Reclassified these workers from "hybrid working" to "commuting" to avoid double-counting; hybrid-working share dropped from ~7–8% to ~2% in re-estimation.
• Added an intrazonal indicator (workplace zone = home zone), improving significance/scale of the workplace logsum coefficient — consistent with a parallel TransLink estimation Joel shared, though smaller in magnitude.
• Group was surprised how many workers report an out-of-home workplace in the same small zone (MGRA); Marty Milkovits noted a similar pattern in a Massachusetts survey; Drew Cooper offered an anecdotal explanation.
• Added industry variables (business services as reference); results mostly intuitive, with a few unexplained anomalies (e.g., military).
• Applied updated models to the full-scale SANDAG model: full-/part-time worker shares matched reasonably; student categories were off, likely because CDAP doesn't control tours by person type — a possible calibration target.
• Notable discrepancy: hybrid-working share drops to ~2% in estimation but rises back to ~7% in full-model application — needs further investigation.
• Remaining deliverable: a memo summarizing the work and documenting survey issues. Joe suggested a more public-facing format (e.g., GitHub issues/user guide) instead of a memo that would live in the project shared drive, to give the work more visibility. The group discussed the similarities / overlap with TfNSW's "home business tours" work and benefit of coordinating with the Travel Survey Working Group to establish a consensus definition for classifying these activities.

Action items:

• Sijia Wang: finalize written memo/documentation of methodology and survey issues.
• Sijia Wang & Joe: coordinate offline on a more public-facing documentation format (e.g., issue-based or user guide).
• Group: aim to wrap up remaining telecommute work within ~6 weeks (by end of September).

Application & Analysis Guide (Joe Flood)

• Drafts are nearly complete; still need a summary section for the telecommuting guide and a quick internal CS review before circulating to the consortium.
• Original presentation date (Aug 25) conflicts with the SANDAG Modeling Forum — moved to September 1.

Action items:

• Joe Flood: finish drafts, complete internal CS review, cross-check content against deliverable requirements, then send to consortium for review.

Telecommute Frequency Model Update (MWCOG/RSG)

• Originally scheduled for Sept 1 alongside the Application Guide presentation; likely to be pushed out further given scheduling conflicts.

Action item: Joe Castiglione to contact Ray to confirm MWCOG/RSG availability and finalize scheduling.

Task Order Amendment Policy / Governance Voting

• Policy proposal was reviewed in last week's partners-only meeting with no major objections.
• A GitHub-based voting process (one vote per member agency) has been set up; invitations were sent via GitHub or email.
• Two pending task order amendments need this policy approved before final sign-off can occur.

Action item: All agency representatives who received a voting invitation — vote on the task order amendment policy or designate an alternate voter as soon as possible.

Phase 12 Issue Template & Prioritization Process

• Draft GitHub issue template (commentable version shared) proposed to standardize issues ahead of Phase 12 prioritization — covering problem statement, affected components, possible solutions, "solution maturity," and testing considerations.
• Debate over including a "level of effort" field: Jeff Newman recommended dropping it — effort estimation is part of firms' competitive process, and AI tools have radically changed effort levels (e.g., tasks once ~300 hours now ~4 hours), making fixed estimates unreliable.
• Broader AI discussion:
  • Jeff Newman: AI dramatically accelerates well-scoped engineering tasks; less impact on exploratory/conceptual work (e.g., deciding how to model telecommute); humans must stay engaged since AI tends to agree with whatever the user proposes.
  • Nicholas Fournier: main challenge is structuring AI work into reviewable chunks, not just estimating effort.
  • Bo Wen: consortium should factor AI-driven productivity gains into work planning to avoid over/under-programming funds; suggested shifting some deliverables-based tasks toward a time-and-materials approach.
  • Joe Castiglione: wants to continue this conversation at a future meeting.

Action items:

• All: review and comment on the draft Phase 12 issue template.
• Joe Castiglione: agendize a follow-up discussion on AI's impact on work planning and prioritization.