The desk's detection systems are continuously busy. They read across a wide field of sources, they find changes and connections, and they identify signals that warrant attention. The volume of what they find is substantial: most of it is routine, some of it matters, and a small portion of it matters substantially. The work of distinguishing among the three is done by people.

Manual review is not a fallback for when the software fails. It is the central element of the work, and the software exists to support it. The systems narrow the field. The team reads what the systems surface, considers the client's circumstances, weighs the practical consequences, and decides what the principal should be told and what should be done. The decision is the part that cannot be automated.

Each review is shaped by context that the systems do not hold. The client's present situation, the political environment in the relevant jurisdiction, the trade press cycle that an item sits inside, the wider pattern of attention that is or is not building, the practical likelihood of an item being noticed and pursued: these are the factors that determine whether a signal is something to act on or something to leave alone.

The team that does the manual review is small by design. The work requires attention, judgement, and a working knowledge of each client's position that does not transfer easily between people. Increasing the headcount would not, in this area, increase the throughput; it would reduce the consistency of judgement that is the central asset.

The review process is, in most cases, conservative. The desk would rather miss an item and have to address it later than raise an item with the client that turns out, on closer examination, not to have warranted attention. The cost of the second error is real: it produces noise in the relationship and erodes the principal's confidence in the work. The cost of the first error is real too, but it is bounded, because the watch continues.

The result, for the principal, is that they hear from the desk only when something genuinely warrants their attention. The volume is low. The signal is high. The watching is the desk's burden, not the client's, and the manual review is what keeps it that way.