Tenets

The non-negotiables. Five principles that guide what the COE does and how it does it.

1. Workflows, not tools

We start with real work that matters to customers and the business. We ask where a workflow is constrained, what is limiting it, and what focused intervention would materially improve it. We don't start with a generic list of AI use cases.

AI fits inside the workflow. It is not the point of the workflow. A cohort that ships a better tool but leaves the workflow structurally the same has not done the job.

2. Venture, not min-max

The COE runs a portfolio of bets, chosen for learning and upside. Not the single highest-ROI initiative identified through analysis.

Some bets will be obvious and grind out durable gains. A smaller number will create disproportionate value. A third kind will fail fast and teach us something. All three are fine. Selecting only for "max expected value" kills the second and third kinds before they start.

Cohort selection reflects this. We prefer exploratory work with curious operators over optimized work with the right-on-paper team.

3. Upstream before downstream

Fix the source of data and work before accelerating what happens with it. Moving bad data faster is a worse outcome than leaving it alone.

When a downstream team is in pain, the temptation is to solve the downstream symptom. The COE's job is to go upstream to the cause. Sometimes the downstream fix is right; often it is not. We commit to the discipline of looking.

4. Collaboration is contribution

Proposals beat comments. Pull requests beat peanut-gallery feedback. Decisions are recorded, not relitigated.

When you want something changed, you propose the change. When you want to disagree, you propose an alternative. When the call gets made, it lives as a decision record. The next person reading this repo can see what got decided and why without asking a meeting to reconvene.

This changes the mental model from spectator to contributor. It is the single biggest shift in how the COE operates over time.

5. Build toward independence

Every cycle leaves Balsam stronger. By the end of the founding period, Balsam runs the COE.

The measure is not whether Goose Group is useful. It is whether the COE keeps working after Goose Group steps back. Each cohort produces trained alumni, documented patterns, and internal infrastructure. Ownership transfers by design, not by accident.

This tenet is the reason we operate in the open, document as we go, and identify the long-term internal owner early.

These tenets are the reason we say no to specific things. They are also how we settle disagreements when two reasonable people see different paths. When the question is "which call do we make?", the answer starts here.