Fleet

The maker does not QA their own change; two independent tests are required. The tester records steps in the issue and moves the work toward release readiness or back for rework.

How it works

The QA owner joins the issue; small failures become separate bugs, larger ones return work to In progress and notify the product group.

Why it is used

Makes test evidence independent from the change author and traceable.

Trade-off

Two separate tests and retesting can lengthen delivery.

Best fit

Product teams with capacity for independent pre-release verification.

Product groups use continuous intake-to-release flow rather than sprint batches; weekly planning examines the board backwards from Ready for release and triages Inbox.

How it works

The planning rhythm is separate from the three-week release boundary; completed work enters the release queue.

Why it is used

Makes near-done work and bottlenecks visible before starting more.

Trade-off

Continuous flow depends on intake prioritization and board discipline.

Best fit

Product groups with continuous intake and regular releases.

The sales handbook ties discovery and scope stages to a recorded use case, budget, dated next step and success criteria approved in writing by the prospect.

How it works

Stage conditions include probabilities and time windows; a demo alone does not qualify as completed discovery.

Why it is used

Replaces assumed sales progress with customer-confirmed context.

Trade-off

CRM fields and written prospect confirmation add work and waiting to the sales cycle.

Best fit

B2B sales teams with multi-stage technical discovery.

Fleet expects engineers to use Claude Code in their daily workflow. In return, no AI-generated work (code, blog posts, handbook edits, customer emails) ships until both author and reviewer have read and understood it line by line; multiline product changes are run locally and the PR is opened from the person’s own GitHub account, not a bot.

How it works

AI output is treated as “cognitive debt”: the author pays it before opening the PR, the reviewer before approving. AI review tools (Copilot, CodeRabbit, Qodo, BugBunny) supplement human review but do not take over accountability.

Why it is used

When writing code gets cheap, the value of review is the guarantee that a human read and understood every line, not the findings; attributing the commit to a person keeps ownership.

Trade-off

Reading every line gives back part of the speed AI adds; the source defines no audit or metric showing the rule is followed.

Best fit

Software teams making AI use expected or widespread while keeping quality accountability with individuals.

Fleet accepts product requests continuously but keeps no feature backlog: a story is prioritized only if it can ship in the upcoming three-week release. Requests are decided at the Feature Fest meeting; a request not prioritized is treated as rejected for now and can be resubmitted. A request must meet at least one of four criteria: a small UX improvement, a company or product maturity goal, a “core” feature, or a “context” feature that one customer considers mission critical.

How it works

The feature-prioritization DRI weighs requests at Feature Fest, and the Head of Product Design triages new ones daily. Core features rank above context features; a context feature can jump ahead only if it is a customer promise or a P0–P2 customer request. The requester is told the decision; prioritized work is delivered or the requester hears about de-prioritization within one business day. Urgent requests can go straight to the Head of Product Design.

Why it is used

To avoid the false promises and staleness of a long backlog, and to give every request a quick, clear yes or “not now”.

Trade-off

The burden of resubmitting a rejected request falls on the requester, and work that does not fit three weeks has to be split to get in. Single-customer features only advance through the exception rule. The page shows no visible date and no measured outcome.

Best fit

Small product organizations with a short, fixed release cycle and a single named decision owner.