KICKOFF · JULY 21 · SAN FRANCISCO

Notes & takeaways.

We met in person with Kim, Emily, and Katrina. Dana was out this week and joins us on July 29. The companion deck from the session is here.

What we decided

Leadership goes first. Kim and Emily take the diagnostic this week. Dana takes it when back, with a Sunday July 27 deadline.
July 29: I present the leadership results and the tool comparison. Leadership sets guard rails and any hard no-gos. Kim is out that week and will comment on materials async.
Leadership acts as the test group for the diagnostic itself. We refine any questions that feel off before the team takes it in Accra.
Accra (Aug 2–6) stays diagnostic-only. Katrina carves one protected block in the retreat agenda.
Tool path: start with one tool for everyone. Consider splitting by team later if needs diverge. That decision lands the week of Aug 17 and holds the rest of the timeline.

The January survey is dated

Leadership's read: the January answers no longer reflect the team. Gemini use has gone up a lot since it landed inside Workspace, and staff were told to prefer it for anything sensitive.
People have since tried Claude and other tools. In January there were zero Claude users; that may not be true anymore, which changes the tool evaluation.
Agreed fix: the team diagnostic in Accra re-asks current tools and current uses. No extra survey, no double-asking.

"Things are moving so quickly, behavior is changing fast. I suspect the Gemini use has gone way up. I'm seeing it in the quality of writing coming across my desk."

HEARD IN THE ROOM

What leadership is seeing in Gemini

Inside the Echidna Workspace, Gemini pulls from internal documents and email, and appears not to draw from the general web. Generic questions came back with Echidna-specific answers.
An old FSG report on the drive could not be surfaced, and general search struggled too. Query technique matters.
Training implication: teach the team how to steer where the tool looks. Internal files versus the open web, and how to phrase queries for each. Goes in the prompt library.

New scope that surfaced in the room

An AI audit of the software stack

We want to know which of your existing subscriptions already have AI inside them: Zoom (the notetaker you can't seem to fully turn off), Grammarly (you have a subscription), Leapsome for performance reviews, Expensify, Bill.com, the PEO, and Vanguard. Emily and I will walk through it together, with input from Kim and Katrina.

Where the data lines sit

HR and finance systems are mostly out of scope. The two things you feel firmly about are compensation data and anything that could enable identity fraud. And writing to a grantee may follow different rules than writing to a funder, so the policy should account for who the audience is.

Human-first or AI-first

An open question for the policy: for anything public-facing, does a human need to draft it first? And is that the same rule for everyone, or does it differ by team?

An onboarding pack

You're expecting roughly six new hires next year, so the training materials will get a light, reusable version for onboarding: the recordings, the prompt library, and the workspace itself as the place to come back to.

Environmental impact, right-sized

The steer from the room was to pay attention without quantifying carbon per token. It matters for how comfortable the team feels, and it won't drive the tool choice.

The champion question is open

Leadership pushed back on a side-of-desk champion. It only works with organic enthusiasm, and the team is stretched. Assigning it without interest or incentives fails.
A dedicated part-time AI role is hard to shape: roughly 30% AI work doesn't pair naturally with any existing function, and a split role is hard to hire for.
Live option: Echidna was already planning a database-focused hire. Accelerating it with an AI mandate is on the table, but the decision feeds the 2027 budget being built right now.
Fallback: external retainer support with set weekly availability. Not real-time, workable.
I owe leadership a short options memo on this. It is time-sensitive because of the budget cycle.

"We have a lot of young mothers here doing another full-time job at home. Oh, by the way, teach yourself in all your spare time. That feels a little unrealistic."

HEARD IN THE ROOM

Later-phase ideas captured (for the 2.0 list)

Program officer dashboards: portfolio snapshots, grant size and length trends, geographic spread.
Automated monthly and quarterly reports from EGGs, and portfolio forecasting as growth flattens.
Research scanning and literature review support, including trends in girls' education language over time, with care not to over-weight US framing.
Writing support for staff working in a second language, relevant to francophone expansion.
Grantee report intake and board prep assembly, as discussed in the deck examples.

Next steps

Kim + Emily: take the diagnostic this week. Dana by Sunday July 27.
Shabnam + Sara: update the diagnostic (re-add current tools and current uses; fix wording flagged in the room) before Dana takes it and before Accra.
Shabnam: leadership results analysis and tool comparison with pricing for July 29. Share materials so Kim can comment async.
Shabnam: champion and staffing options memo for Kim and Emily, timed to the budget cycle.
Emily + Shabnam: schedule the software AI audit walkthrough.
Katrina: work through the timeline and scheduling guide; the Accra diagnostic block is the priority hold.

Prepared by SA Advisory from the July 21 kickoff session. Corrections welcome → aggarwal.shabnam@gmail.com