Sentinel monitors public advertising and news streams, scores every item against interest profiles your analysts author, and surfaces matches to a human triage queue. The AI must quote the exact evidence behind every verdict — and it never acts on its own.
Two modules, one platform
Both modules share one discipline: ingest a public stream, score each item against analyst-authored interest profiles, and put every judgment in front of a human.
Watches public ad-transparency sources and scores every advertisement against your interest profiles — so launches, claims, and campaigns that matter to you never slip past.
Ingests licensed and open news, collapses syndicated copies into a single story, and scores each story against your profiles — with source licensing respected at an architectural choke point.
Trust architecture
Enterprise intelligence fails when you can't explain why the system said what it said. Sentinel is built so that every claim is checkable and every action traceable.
The judge must quote the verbatim text that justifies its verdict — and the quote is verified to exist before the score counts.
Sentinel never acts on an ad, an advertiser, or a story. Non-action is enforced by egress zoning in the architecture itself.
Every decision lands in an append-only, hash-chained log. What the system judged, and why, is replayable end to end.
Workspace-scoped row-level security separates every tenant's data — and when in doubt, access is denied by default.
Who it's for
Sentinel serves teams whose judgments carry weight — where "the AI flagged it" is not an answer, but "here is the exact evidence" is.
Track competitor launches, claims, and campaigns as they surface in public streams.
Catch misuse of your brand and lookalike claims early, with cited proof attached.
Watch regulated categories with an audit trail that stands up to scrutiny.
One deduplicated story stream, scored against exactly what your desk cares about.
Early-access partnerships are open for intelligence, compliance, and risk teams.