Industry
Recruitment & staffing
A chain of external offices and no single view of what is late
The problem
Cross-border recruitment runs on a chain of external offices, each with its own paperwork, timezone and reporting habit. A coordinator cannot say which cases are late, which arrivals are due this week, or which need intervention today, because the answer lives in a different place for every office.
This is an operations problem before it is an AI problem, and treating it as an AI problem first is the common and expensive mistake: a summarising assistant over a case model nobody trusts produces confident summaries of the wrong state.
What we built
A daily operations and decision board over one case model spanning every partner office and source country — stage, SLA state, delay detection, arrivals on a 72-hour horizon, and an exception centre listing the cases that need a human decision now.
Read the Recruitment Operations case studyWhat transfers
The case model has to survive every office
Seven countries' processes do not agree, and a model that encodes the most common one forces the rest into exceptions that then have to be tracked outside the system. Getting this right is most of the work and none of the demo.
Exceptions, not totals
A board reporting how much work exists is a number. One saying which cases are late, which need a decision today and which are about to breach is a tool. The value is in the filtering, and the filtering is the hard part.
The application owns the state
Stage, delay and SLA state are computed from recorded facts, which is what makes them auditable and what makes it possible to defend the claim that a case is late. A generated summary on top is useful; a generated summary that *is* the state makes every counter unciteable.
Capability: AI Agents & Automation