What you're looking at, who built it, and what happens to your data.

BEACON is the AI-native executive canvas — a single, live, structured view of strategy, people, and roadmap. Edited through chat. Reportable anywhere. Built so the operating model stays a living document instead of a stale slide deck.

BEACON is one of four products in the SenseAI family. The other three (TIME, SIGNAL, DUNBAR) are on the roadmap; click the B in the top-left for the family overview.

XCeedance is using BEACON to author and share weekly status updates against the company's North Star, objectives, and key results. The page you're reading right now is generated from the same operating record an exec would edit through chat — so the report and the underlying model never drift.

BEACON is a SenseAI product. SenseAI is the developer; XCeedance is the user. No XCeedance-internal decisions are made by SenseAI; BEACON simply renders what XCeedance authors.

Two layers, with deliberately different lifetimes:

• Authored content — the North Star, objectives, key results, narrative documents, and review confirmations that XCeedance leaders write. Stored in a Postgres database, encrypted at rest, accessible only to authenticated XCeedance users via the /api/workspace/ endpoint.
• Visitor session state — what you've clicked into during this browsing session, your unlock progress, the welcome splash. Held in sessionStorage on your browser; cleared automatically when you close the tab. Never transmitted anywhere.

When XCeedance edits content via the Make Edits chat (the ⌘K dock), the prompt + the relevant workspace context are sent to Anthropic's API to produce a structured proposal. The proposal is shown as a diff and applied only when a human accepts it — nothing the AI suggests commits silently.

BEACON's Anthropic integration runs under a zero data retention agreement: prompts and responses are not used to train models and are deleted after the response is generated. Customers who prefer their own AI engine can route through an MCP server they control instead of Anthropic; in that mode no user-authored content reaches BEACON's vendors at all.

When the AI proposes an edit to something a human reviewed in the last 7 days, BEACON requires the human to acknowledge an override checkbox before the change can land. The machine does not get to silently overwrite recent human decisions.

Read-only visitors (this view) see only the published snapshot. Edits require authentication and run through XCeedance's workspace key. Per-user role permissions and field-level version history are on the near roadmap; until then, every edit is timestamped and attributable to the human who accepted it.

Each cadence layer (objective → quarterly review, KR → monthly review) carries its own freshness signal. When something is overdue, it surfaces in the River for a human to confirm or edit. So the data on this page either is fresh or is visibly stale — never silently both.

BEACON uses PostHog for product analytics — to know which surfaces are getting traction and where visitors get stuck. Captured: page views, click events on key surfaces (Make Edits opened, KPI clicked, Sheet opened, review confirmed), and the visitor identifier you provided in the welcome splash.

Not captured: the contents of authored prose (objective titles, KR text, narrative bodies). Autocapture never includes input values. The PostHog instance runs under a separate, isolated tenant from any other XCeedance system.

Don't want analytics on at all? Removing theNEXT_PUBLIC_POSTHOG_KEYenv var disables tracking entirely — the whole module no-ops.

For BEACON product questions: SenseAI · hello@senseai.example.

For XCeedance content questions: the owner named on each objective on the canvas. Click any objective to see the DRI and the linked teams.

For data security or privacy concerns: we're happy to walk through the architecture in detail — the encryption boundaries, the AI-engine routing options, the retention model.