What is a detection?
A detection is something a Digital FTE noticed in the course of its work that its manager may want to act on, raised in the morning brief with the evidence behind it and, where the FTE can propose one, a suggested response that waits for approval. It is the mechanism by which an AI employee behaves like a colleague rather than a tool: a tool does what it is told, a colleague tells you what they saw.
Detections are not commands and not alarms. They are observations ranked by the FTE's judgement of whether they matter, with the evidence so the manager can judge for themselves.
What do detections look like per role?
A Finance Controller FTE detects possible duplicate payments, unusual amounts, timing shifts, and a supplier whose invoice format changed. A Search Visibility FTE detects a competitor page that began outranking yours, a page that lost its citation, and a new question appearing in the prompt samples. A Customer Support FTE detects a spike in one category, a knowledge base article customers keep misreading, and a product change generating confusion.
A Performance Ads FTE detects a spend spike, a platform policy flag, and a performance collapse on one creative. A Market-Entry Strategist FTE detects a named risk signal moving, such as a regulatory update. A Recruiter FTE detects a candidate source drying up.
- Anomalies: duplicates, unusual amounts, spikes
- Changes: a competitor, a platform policy, a format
- Opportunities: a new question asked, a gap a page could fill
- Risks: a named signal from a plan moving
How does a detection work on ftes.ai?
On ftes.ai detections are produced by the roles' measurement and monitoring jobs and by the FTE's reading of the data it works in. Each is stored with its evidence, surfaced in the Today tab's detections block after the waiting-on-you items, and logged. Where the FTE can propose a response, such as pausing a creative, correcting an article, or drafting a page, the proposal appears in Approvals with the detection as its reason.
The manager can act, dismiss with a reason, or ask the FTE to watch. Dismissals are logged, and a detection the manager dismisses repeatedly is a signal to adjust what the role watches for.
Why are detections separate from waiting-on-you?
Because they are optional. Waiting-on-you items block the work: an unapproved proposal does not happen. Detections do not block anything; they inform. Putting them after the decisions in the brief means a manager with two minutes clears what blocks the day, and a manager with ten reads what the FTE saw.
A detection with a proposed response becomes a waiting item once the FTE has drafted the response, so the two blocks connect: the detection explains why the proposal exists.
What is a detection often confused with?
It is confused with an alert, which is a threshold crossing with no judgement; a detection includes the FTE's reading of whether it matters and why. It is confused with a proposal; a proposal is an action awaiting approval, while a detection is the observation that may lead to one. And it is confused with an incident; most detections are ordinary, and the incident plan is for the rare bad day.
It is sometimes expected to be exhaustive. It is curated by the FTE's judgement, and the monthly review is where the manager checks that judgement against what actually mattered.
Where do you see detections in the workspace?
In the Today tab's detections block each morning, in the Detections view with history and evidence, and in the audit trail with the manager's action or dismissal. Patterns across a quarter are part of the monthly review and often become remit rules or watch-list changes.
Related terms: morning brief, proposal, mission (diagnostic), targets, incident.
In the first month, act on detections sparingly and dismiss with reasons generously. The reasons teach the role what the manager considers worth raising, and by the second month the detections block is short, specific, and usually right. A manager who never dismisses trains nothing; one who dismisses with a sentence each time ends up with a colleague whose observations they trust.