FAQ

The questions people keep asking. If yours isn't here, open an issue.

Time commitment

How much time does being in a cohort take?

Depends on your role. Core owners commit ~50% of their time during an active cycle (4–6 weeks). Participants ~20%. Adjacent ring colleagues ~5% (office hours).

See roles-and-time.md for the full picture.

Is this extra work on top of my job?

For core owners and participants, no — the cohort is how you do your existing work for a few weeks, not a side project. If you're doing product data work, a product data cohort is your product data work done differently. Adjacent ring participation is light enough to fit alongside normal responsibilities.

What happens after the cycle ends?

Core owners graduate into alumni status. They stay available for office hours, coach the next cohort, and carry the learning into their day job. The 50% commitment is not permanent.

Scope and prioritization

Who decides which cohorts run?

The Operating Council — a subset of Balsam's Leadership Team. They meet monthly, review the backlog, and pick the next cohort. The COE presents the options and the reasoning.

The AI Council is not the prioritization body. Their role is enablement and standards — tooling, policy, shared infrastructure across AI efforts. They get COE readouts but don't pick cohorts.

See governance.md.

How many cohorts run at once?

One to start. The first cohort (Product Data, launching May 2026) runs solo. Cohort 2 begins in Month 2–3 once the rhythm is set. From Month 3 onward, we may run two concurrent cohorts depending on capacity.

Why Product Data first?

Three reasons. One: the strategy sprint already mapped it in detail — 19 steps, 18 bottlenecks, strong candidates for the cohort. Two: Balsam decided not to move forward with a PLM this year, which makes the existing workflow carry more weight. Three: Product Data is upstream of retailer onboarding, content production, and international expansion — fixing the source helps more than fixing the downstream symptoms.

Is marketing in the first cohort?

No. Marketing is ripe for reimagining, and Pascal Staud's introduction to Karima is happening separately. Marketing moves as a parallel engagement rather than a COE cohort during the founding period.

Deliverables

What does a cohort actually produce?

Depends on the cohort. Typically:

A workflow map of how the work moves today.
Prototypes, scripts, or internal tools that remove friction.
Skills (reusable AI prompts) and playbooks.
A readout for the Operating Council on what was built and what moves next.
A handoff package if the work needs production engineering.

See how-cohorts-work.md for the full rundown.

What should I expect by the end of Month 1, 2, 4?

Month 1: COE launched, Cohort 1 roster confirmed, workflow mapped, first prototype shipped.

Month 2: Cohort 1 complete, readout published, Cohort 2 starting.

Month 4: Three cohort cycles done, three mapped workflows, growing repository, first alumni coaching.

See the founding brief for detail.

Build capacity

Who builds the production version of what the cohort prototypes?

Depends on the work. Options:

Balsam's internal pods (France's team has committed 20% capacity).
External build partners — Spatialedge is a strong candidate for data engineering work.
Technology vendors directly (Snowflake, Databricks, Anthropic) where it fits.
Goose Group contractors for narrow short projects.

The COE scopes the work and manages the partner engagement. Balsam funds it from normal business budget, not from the COE retainer.

See build-partner.md.

What's the budget for external engineering?

Depends on the project. A focused Spatialedge engagement for Balsam-shaped work is likely in the $80–150k range, scoped as a 6–12 week project.

Could this engineering spend have gone to the PLM instead?

The Operating Council is budgeting holistically: COE retainer + expected partner engagements against what PLM implementation and licensing would have cost. The early read: comparable annual investment, very different shape of outcome.

Tools and infrastructure

What's the right AI tool — ChatGPT or Claude?

Both. The cost difference is small and the models are converging. What matters more is that knowledge is portable between tools, not locked inside one team's workspace. The knowledge repository the COE builds solves this — learnings live in a shared place regardless of which tool generated them.

The AI Council can refine the standard over time. For now: use what works.

Why GitHub?

Because it gives us collaboration-as-contribution. Proposing a change (pull request) beats commenting on a doc. Decisions live as commits. The record of what got decided and why is legible without asking someone to dig through Slack.

Cohort members will learn the basics as they need them. It's not a prerequisite — it's a tool that rewards the effort.

Governance

What's the difference between the Operating Council and the AI Council?

Operating Council: prioritizes COE work. Close to the business. Picks cohorts. Funds initiatives. Recently formed subset of the LT.
AI Council: enablement and standards. Tooling, policy, training. Gets COE readouts but doesn't prioritize cohorts.

Will the COE coordinate every AI effort at Balsam?

During the founding period (Months 1–4), Goose Group acts as connective tissue across the broader AI landscape — Pablo's initiative, Bernie's efficiency work, training programs — so they see each other. We don't run them. We make sure the right conversations happen.

By Month 5, this coordination function transitions to an internal owner (likely emerging from the AI Council or a recovered PM).

See governance.md.

Who's the long-term owner of the COE?

Not named yet. By Month 3–4, a candidate will have emerged. Criteria: credibility across functions, time to take it on, curiosity to keep it sharp. Claire and France make the call on Goose Group's recommendation.

Getting involved

How do I suggest an idea?

Use the intake template. Open a pull request or email the COE operating team. Every submission gets a response within a week.

Can I join a cohort?

If your work touches a cohort's workflow and you can commit the time, yes. Talk to the cohort sponsor or the COE operating team. See how-cohorts-work.md for what kinds of people cohorts need.

Where do I follow what's happening?

This repo — read the readouts, browse cohort packages.
Monthly readout meetings (open to the whole company once the rhythm is set).
Slack channel — #balsam-coe (to be created).

Still have questions?

Open an issue on this repo. Or ping Mike and Alex directly.