Sod first. Landscaping later.

Project kickoff · with Kim, Dana, Emily & Katrina · San Francisco · July 21, 2026

A web companion to the kickoff conversation — with quick examples you can click into along the way.

THE STEER

What you asked for.

Setting foundation before exploring use cases. Your board pulled the scope back from designed use cases to setting foundation, which I think makes sense.

A safe place to experiment. Enterprise accounts, clear rules and policies for your data and team, and making space for the AI workflows to grow and mold where they naturally fit.

Internal only. This is about how Echidna works — not AI in your portfolio, and nothing kid-facing.

‟An architectural firm that planted sod and waited to see where people walked — before landscaping.”

KIM — ON WHY USE CASES WAIT

THE PLAN

Four months in three phases.

The current picture

JULY — AUGUST

I start with the four of you, then the full team comes in in Accra (Aug 2–6), which gives us everyone's POV. We'll get a deep read of where Echidna's team actually is today.

The secure foundation

AUGUST — SEPTEMBER

We'll pick a tool together. Your IT partner will stand up the sandbox to my specs. The policy and updation cadence comes into place, plus the data rules and risk framework.

Education & handover

SEPTEMBER — OCTOBER

Training built around role and comfort level on the tool and policy we chose. We'll also set up the go/no-go framework for evaluating new tools, run the endline diagnostic, and hand everything over.

Next session — July 29: we review your leadership diagnostic results together, plus a pros-and-cons overview of the top AI tools we're currently considering.

▸QUICK EXAMPLEWhat a go/no-go framework looks like in practice

Say someone on the team wants to try a new AI note-taker. Instead of a debate, the framework asks four questions:

What data does it touch — and what zone is that data in?
What kind of account would we run it on — enterprise terms, or consumer?
Does it stay on by default, or only work when invoked?
Who owns the decision to keep, restrict, or remove it after a trial?

Clear answers to all four — the trial can go ahead. A red flag on any — it comes to the working group first.

WHAT WE ALREADY KNOW

Based on your January survey.

Twelve responses across the team. Comfort spreads from 4-10 — but comfort and use don't line up: your most confident people use AI least, and two of the three daily users rate themselves lowest. Four things hold across the board:

Starting from fear: everyone's worried about the data implication

Grantee, partner, and staff data concerns came up over and over. Several of your staff refuse to put Echidna documents into AI at all today.

‟Don't over-trust it”

The second theme: accuracy. Today's use is almost all text work; drafting, editing, and summarizing (9 of 12). Analysis: just 4 of 12 — and no one has touched agents or automations for patterned work (0 of 12).

ChatGPT + Gemini bias

People really like ChatGPT and use that and Gemini primarily. Any consolidation/change will carry a switching cost.

Caution that comes from values

Many people hold back AI use with deep intention — to protect their own writing, their judgment, the human touch. 3 people raised water and energy use concerns.

THE 2×2 THAT FALLS OUT

Plotting comfort against actual use, four groups emerge, and each needs something different.

HIGH COMFORT · OCCASIONAL

Confident but idle

Rated 9-10, use AI only occasionally. They have the skills already. What's missing is a reason and a use case worth their time.

HIGH COMFORT · DAILY

The power user

One person, ~10 task types daily. Give them stretch material, and lean on them as a peer resource for the team.

MID COMFORT · WEEKLY

The cautious middle

Most of the team: text work a few times a week. Ready to move with clear rules and a shared baseline.

LOW COMFORT · DAILY

Daily but unsure

Rated 4-5, yet use AI every day. They feel least equipped while relying on it most. Confidence and guardrails come first for this group.

▸QUICK EXAMPLEWhat ‟agents and automations for patterned work” could mean here

Beyond drafting, AI can take on recurring, patterned work end to end — always with a person reviewing the output. Two examples shaped to Echidna:

Grantee report intake. When a grantee report lands, an automation reads it, pulls the updates against the grant's stated milestones, and drafts a one-page summary for the program lead to review — the same steps you do by hand today, every reporting cycle.

Board-prep assembly. An agent gathers the quarter's portfolio notes, flags what changed since the last board meeting, and assembles a first-pass briefing pack — hours of collation compressed into a review job.

Writing support. Polishing drafts for team members writing in a second language — increasingly relevant as you look toward francophone Africa. The thinking stays theirs; AI helps the words land.

WHAT THE CONSENSUS HIDES

Where the mismatch lies today.

Intent

Practice

‟We protect confidential data.”

AI subscriptions are personal and free. Your Workspace is enterprise — but the AI tools around it (ChatGPT, Grammarly) mostly run on personal / free accounts that Echidna can't control.

‟Nothing sensitive goes in.”

Always-on tools don't ask for permission. Grammarly reads everything typed in where it's applied; Gemini reads emails and docs, and listens to calls, which means sensitive data may be going into AI servers without realizing it.

‟We all draw the same line.”

Everyone's line may be different. Some pay to opt out of AI training. Some never upload anything. Others avoid ‟confidential info generally”, which are different lines.

‟More concern means more caution.”

Concern tracks inversely with use. The most worried staff use AI least. For roughly a third of the team, the policy is what opens the door to using AI at all.

Goal: get everyone onto one secure enterprise tool, and deal with the shadow tools — govern them or pull them out.

▸QUICK EXAMPLESee what's ‟always on” in your own account, right now

In Gmail, open Settings → General → Smart features. That toggle governs whether Google's features (including Gemini-powered ones) can read your email and docs to personalize what you see. Most people have never looked at it.

Your Google Workspace itself is on Echidna's enterprise accounts, and Google's privacy commitments there are real: workspace.google.com/security/ai-privacy. The gap is the AI tools around it — ChatGPT and Grammarly running on personal or free accounts, where none of those protections apply.

Same person, same work, completely different rules depending on the tool — that's exactly what the diagnostic will map.

WHAT WE ASK NEXT

The next diagnostic digs deeper.

ALREADY KNOWN

Comfort ratingsWhich tools people useBroad concernsFrequency of use

DOUBLE CLICKS

What kind of account is it, really? — personal, free, work, or enterprise, tool by tool What's already reading your files — the always-on tools with access to email and documents can seem innocuous, but aren't Red / yellow / green flags — each Echidna data type, sorted by privacy indicator Where the line is — where AI helps your team's judgment vs. replaces it, and whether cost and environment matter for the team The wish list — what the team wishes AI could do for them that it doesn't today

▸QUICK EXAMPLEWhat a red / yellow / green data sort looks like

Each Echidna data type gets a zone, and each zone gets a rule. A sketch — the real version comes out of the diagnostic together:

REDGrantee financials, staff HR records, unannounced grant decisions — never into any AI tool, enterprise or not. YELLOWInternal memos, board drafts, portfolio notes — enterprise tool only, on a work account. GREENPublished reports, public research, your own drafting — any approved tool.

The point isn't the categories — it's that everyone sorts the same data the same way, without stopping to wonder.

DEFINITION OF SUCCESS

What success looks like.

1 · Current state & posture

‟We agree on what's true about Echidna today, and where the red lines/non-negotiables are.”

2 · Provider recommendation

‟One tool, chosen deliberately, that everyone's AI work lives on.”

3 · Secure sandbox

‟People feel comfortable to try AI without fear.”

4 · Policy & data governance

"Echidna feels like a workplace with clear AI policy and steer for the staff to increase productivity without sacrificing privacy"

5 · Risk framework

"It is clear what the risks are when introducing specific AI tools and workflows."

6 · Education & training

"Echidna staff feel more comfortable and competent to use AI to support them in their day to day workflows."

▸QUICK EXAMPLEWhat a ‟living policy” page could look like

Not a 20-page PDF — a short internal page anyone can check in ten seconds. An excerpt might read:

‟Use your Echidna account, not a personal one, for anything work-related.” ‟Green-zone data: go ahead. Yellow: enterprise tool only. Red: never — ask if unsure.” ‟New tool you want to try? Run it through the go/no-go check first.” ‟Last updated September 2026 · Owner: [AI champion] · Questions → #ai-questions”

The ‟living” part: it has an owner, an update cadence, and a place to ask — so it changes as your use does.

YOUR SETUP TODAY

How do you work today?

Before any tool decision, I'd like to spend time mapping what already exists — so the AI tool we choose supports the workflows you actually run.

The tooling you use

What's in the stack today — Google workspace, grants management, finance, communications, etc.

How it's set up

Where files live, how sharing works, who administers what, and what your IT partner covers vs what falls to the team to decide.

The workflows that matter most

Grant cycles, board prep, portfolio reviews, grantee communications — the recurring work an AI tool would need to support, and where the friction sits today (if any).

OPEN QUESTIONS

Four areas to push on together.

AI champion

It would be ideal to identify, soon, an in-house AI champion — someone who can work closely with me to imbibe the learnings, understand and champion staff needs, and think through use cases once we're done. Who might that be?

Workspace understanding

Together with the AI champion, I'd like to map how your Workspace is actually set up — file storage, sharing settings, who can see what. An AI tool inherits every folder shared too broadly, so we should understand it before the tool does. Staff are already asking concrete questions here — like how to block external AI notetakers from joining Echidna calls.

Environmental impact

Three of your staff raised water and energy use concerns — how deeply do you want me to research and evaluate tools against this concern?

Handover support

Once these sessions end in October, the AI learning shouldn't stop. What handover plan should I put in place to ensure you're hitting the ground running post-October?

Does this track? What might you add?
What does success look like from where you sit?

ECHIDNA × SA ADVISORY · KICKOFF · JULY 2026