Competitive intel rots faster than any other internal document. The brief written for the January sales kickoff is confidently wrong by March — pricing changed, the competitor shipped the gap feature, two of the win/loss patterns reversed. Sales keeps quoting it because nothing replaced it.
The underlying notes often exist — call notes, win/loss writeups, analyst clippings — but synthesis is nobody's Tuesday job, so it happens annually and ages daily.
A scheduled Fleet workflow re-synthesizes the brief monthly from your accumulated notes corpus, a review step flags claims that have gone stale or lack a source note, and the PMM approves before the team is notified.
genflows:
- name: competitor-brief
schedule: "0 9 1 * *" # monthly
steps:
- {name: synthesize, prompt: "Synthesize the competitive brief from the notes corpus: positioning, pricing, recent moves, our win/loss patterns. Date every claim to its newest source.", corpus: ["intel/notes/**/*.md", "intel/win-loss/*.md"], kind: report, out: brief.md}
- {name: staleness, prompt: "Flag claims older than 90 days or supported by a single source. Flag anything sales would repeat to a prospect that we can't back.", depends_on: [synthesize], kind: review, out: flags.md}
- {name: pmm-ok, depends_on: [synthesize, staleness], kind: approval, out: decision.md}
- {name: announce, depends_on: [pmm-ok, synthesize], kind: notify, out: sent.md}
The notes corpus is append-only and cheap to maintain — drop a call note in, the next monthly run incorporates it. Synthesis stops being a project and becomes a side effect.
Append notes continuously → scheduled monthly synthesis → staleness review → PMM approval → notify the field. The brief's freshness is bounded by the schedule, not by someone's spare time.
Product marketing and product teams that already capture competitive notes but synthesize them rarely — anywhere sales is working from a brief older than the competitor's last release.
No. The corpus is your own notes — call writeups, win/loss docs, clippings your team saves. Fleet's workflow engine synthesizes and governs what you've collected; document steps run tool-free, so the draft can't wander off to browse.
The schedule (it happens without anyone remembering), the staleness review (claims get dated and flagged), and the gate (PMM approves what the field hears). The chatbot gives you a one-off summary; the workflow gives you a maintained, governed document.
One binary. Five minutes. See every agent, coordinate every handoff, and keep a full audit trail of what your fleet did.