One goal. A team with work to do.

You can authorize internal research, drafting and reporting within a saved project and its limits. A server routine can request that work without your browser staying open. Publishing, sending, live website changes and spending need separate approval in a supported workflow.

No. Kate coordinates a proposed team drawn from twelve specialists. Review the selected roles, tasks and credit estimate before authorizing work. Channel teams share specialists and a common brief.

No. Choosing a channel scopes the internal work. External delivery needs the relevant connection, supported destination, exact content and approval. Some outputs are recommendations or drafts for you to implement.

Use the project controls to pause or resume future internal work and review its saved limits. A pause cannot undo work that was already admitted. Open the linked mission to check work already queued or running.

It links to the exact saved deliverable versions from your selected day, week or month. Current tasks and next planned runs are shown separately from that historical work. Counts describe recorded internal work, not approval, publication or business results. Open saved report history for the reports recorded at the time.

Open Continue this project with Kate and ask for a next brief using its saved work. The proposal costs one credit. Review the suggested changes before choosing to apply them. Applying updates the brief for future work and preserves the project's team, routine, permission and limits; it does not request a run. An already authorized routine can use the updated brief later.

Set simultaneous-run capacity and high, normal or low priorities for this business. See active work and projects waiting for review or a slot. Existing runs, including pending reviews and unknown outcomes, keep their slots. Saving capacity and priorities does not activate projects or grant internal-work permission. Queue order is not a promised start time.

If a run is held, inspect its recorded reason. Also check the project permission, limits and existing work: a request that was not admitted may not create a run. A scheduled time is not proof of delivery. Model and connection failures are not replaced with invented results.

The project permits one active run at a time. Pending Kate acceptance, a required answer or unfinished work can keep that run open and hold the next routine. Open the current mission and resolve its review or question. Accept only work you have inspected; acceptance of Kate's summary does not approve every source or publish anything.

Open View delivery coverage to see which specialists delivered, could not finish, stopped or are still pending. Kate's summary can be partial. Accepting that summary applies to its saved version only; each underlying draft still needs its own review.

Grace records an AI assessment of each required saved report and supported social row. The server checks that those assessments refer to the expected versions and identifies missing, stale or duplicate entries. This establishes which work was assessed, not that every factual judgement is correct. Inspect the findings and original copy before accepting or using it.

Open Preview social drafts on Zoe's current report, inspect the exact copy and choose Save selected copy as drafts. Eligible rows become editable, unscheduled drafts in Calendar. Add any missing image and complete the separate review and delivery steps. Unsupported formats and unresolved placeholders stay held. Check business facts before using the draft.

Open Delivery and observations for the saved draft. Record an owner-reported count, measurement window, time zone and source, or choose no data. A measured zero is different from no data. Your observation is associated with that draft; it does not verify publication or prove the copy caused the result.

A later authorized run saves up to ten eligible earlier observations as context for its planning. That set stays fixed for the run. Saving an observation does not start work or change existing learning decisions. Review the next output to see whether the evidence supports its recommendations.

No. Department specialists use the saved brief, owner-provided business knowledge and the run's evidence. They do not read or add to the shared store of unreviewed model notes. Saved learning decisions and observations keep their own review status and sources; their presence is not proof that a later recommendation used them correctly.

Open View model calls and reported cost on a run. It shows recorded phases, outcomes and available provider-reported cost for the displayed calls. Unknown cost stays unknown. The subtotal is not the full run cost and may exclude other services or unrecorded retries; it is separate from your credit allowance.

Reports show recorded department activity and its coverage. Delivered work is not proof of publication or a business result. Business outcomes are not measured in these totals; missing evidence stays unknown. Owner-reported observations and model-call cost records remain separate evidence.