Stack cards turn the directory into a decision product: which combination to try, where it fits, and what to verify before rollout.
Engineering teams moving from autocomplete to human-reviewed multi-step agent work.
Risk: The main risk is unmanaged agent spend and weak approval policy around repo-changing work.
Operators and product/marketing leaders who need shared context without buying every user an IDE agent.
Risk: Connector scope and workspace data controls matter more than model preference.
Teams whose work objects already live inside Atlassian Cloud.
Risk: Rovo credits pool at the organization level, reset monthly, and do not roll over; model agent, reasoning, and Teamwork Graph demand before broad rollout.
Product teams validating a UI direction before full implementation.
Risk: The stack fails when there is no visual target or the target is not checked against a rendered viewport.
Support teams that need grounded answers from product, help-center, and policy content.
Risk: Risk rises when source content is stale or the bot can answer beyond approved policy.
Repo-changing work, research synthesis, and repeated QA where success needs evidence instead of a single answer.
Risk: The stack needs a hard stop condition, budget ceiling, and human approval for deploy, data, or permission changes.
