A program manager spends Monday morning reconciling four trackers into one slide. The agent can do the reconciliation. It cannot decide what the executives need to hear. Status synthesis works when the agent reads Asana, Linear, Jira, and Monday in parallel, drafts a candidate brief with citations to every ticket it touched, and a named PMO reviewer approves the narrative line by line. The artifact that reaches the leadership channel carries the reviewer's name, not the agent's. See Dock for project management for the broader pillar.
Asana, Linear, Jira, and Monday stay the system of record for the raw data. Dock is the system of record for what the AGENT INTERPRETS. Each Dock row carries a pointer back to the platform record, agent identity, decision, reviewer, and timestamp. The agent re-fetches platform data via fresh API reads when it needs current state.
The surface: a Weekly Status Synthesis table
| Initiative | Source tickets | Agent draft line | Risk flag | Reviewer | Status |
|---|---|---|---|---|---|
| Billing migration | JIRA-4821, LIN-902 | "Cutover slipped one sprint; auth dependency on platform team" | Amber | priya@ (PMO) | Approved 05-29 |
| Mobile redesign | ASANA-7733, MON-114 | "Beta in 12 accounts; two P1 bugs open, one regression" | Red | priya@ (PMO) | Edited then approved 05-29 |
| Pricing experiment | LIN-1044 | "Experiment 60 percent ramped; results read on 06-12" | Green | priya@ (PMO) | Approved 05-29 |
The pointer column is not decoration. A reader can click JIRA-4821, see live status, and confirm the agent did not paraphrase a stale field. The agent collaboration primer explains why the pointer matters more than the prose.
The workflow
Every Monday at 0700, the synthesis agent runs. It queries the four trackers for tickets tagged exec-visible, normalises status fields, and groups them by initiative. For each initiative it drafts a one-line narrative and assigns a risk colour. It writes the candidate rows to the Dock table and pings the on-duty PMO reviewer. The reviewer reads each row, clicks through to the source ticket on anything ambiguous, edits the prose where the agent overstated or understated, and approves. The approved table renders into a Notion page and a Slack digest. The agent never posts to the exec channel directly. Identity is enforced upstream; see agent identity.
Why it matters
The Standish Group has documented for two decades that most large projects miss scope, time, or budget targets (Standish Group). PMI's Pulse of the Profession surveys consistently find that organisations waste a meaningful share of every project dollar to poor performance (PMI). Status reporting is where that waste becomes visible or stays hidden. An agent that drafts honestly and a reviewer who signs the narrative produces a paper trail leadership can trust. The same architectural pattern applies to research and to the product function.
Try it
Point Dock at your four trackers. Run one synthesis cycle next Monday. Read the agent identity primer first so the reviewer signature is real.
FAQ
Who owns the brief if the agent drafted it? The reviewer. The agent produced a candidate. The PMO signature on the row is the publication event. If the brief is wrong, the reviewer is accountable, and the row history shows what the agent proposed before the edit.
What stops the agent from posting stale data? The agent re-fetches from each tracker at run time. It does not read a cached Dock copy. The Dock row holds the interpretation, not the source state. If Jira changes at 0701, next Monday's run reflects it.
Why not let each tool's native AI summarise its own slice? Because the exec needs one narrative across four tools, and each native summariser only sees its own data. Cross-tool synthesis is the job. Single-tool summaries are not.
What if the reviewer disagrees with the agent's risk colour? They edit the row. The history records both the agent's original colour and the reviewer's correction. Over time you can audit the gap and tune the prompt or the threshold.
