This case study is drawn from Mercer's official public materials on global human capital management vendor selection. It is an independent teaching simulation, not real client correspondence or real client results, and Mercer has not used, reviewed, endorsed, sponsored, certified, or commissioned Minerva Advisor. Before Minerva ever saw the decision, we removed the case's later-stage answer and rebuilt only the facts that were knowable at the moment of decision into a training dataset. What follows is an actual replay of that Decision Room run, showing exactly what the system produced under this proof boundary.

The case: a global car rental organization operating three brands across roughly ten thousand locations and more than thirty thousand employees needed to replace fragmented human capital management systems outside North America. Requirements were complex and varied across regions, internal consensus was lacking, and a traditional request for proposal process was seen as slow, costly, and not guaranteed to identify the best-fit solution. The question put to Minerva was not which vendor to pick, but whether the team should launch a full RFP immediately or first run a time-boxed, stoppable fast evidence screen to test whether a shared baseline of requirements could even be established.

Minerva's run produced a documented work receipt rather than a polished summary. It drew on one public source, identified four distinct roles in the decision, separated five verified facts from two inferences and three open unknowns, compared two possible paths forward, and preserved three alternative explanations along with one explicit reversal condition. Every one of these items is available to the executive as a checkable artifact, not a narrative conclusion.

What was known: regional business, systems, regulatory, and integration requirements were numerous and varied, a traditional RFP would be lengthy and expensive, and no credible comparison standard yet existed. What remained unknown: which requirements form a true global common baseline, whether regional divergence stems mainly from regulation or from preference, and how well any candidate product would fit on cost, migration effort, and long-term flexibility. Minerva did not resolve these unknowns for the executive; it kept them visible as open items requiring further evidence.

The strongest challenge came from the advisor tasked with stress-testing regional risk: the fast screen's advantage only holds if a credible global baseline and regional-exception standard can be converged on within the time-box. If that convergence fails, or if the screen surfaces credible evidence that regional must-have requirements remain unmet, the low-cost, reversible rationale collapses. The explicit reversal condition Minerva attached to its recommendation was this: if the time-boxed screen cannot resolve the baseline question within its deadline, or if major regulatory or integration gaps surface, escalate immediately to a full RFP.

Mercer's later-stage case material describes a formal On-Demand Solution Evaluation process, an automated requirements questionnaire, product research, dashboards, drill-down comparisons, a business case with return-on-investment figures, executive sign-off, and the vendor ultimately selected. None of that was given to Minerva. It was deliberately held out so the test would show whether the system could design a sound, stoppable path to a vendor decision using only what was knowable before that later-stage evidence existed.

Three advisors reviewed the same judgment from different angles. One focused on whether comparison criteria could realistically converge across regions. A second modeled the consequences for regional HR teams, technology staff, procurement, and decision-makers. A third specifically tested whether the fast path might miss essential regional requirements. All three independently converged on the same recommendation: run the time-boxed evidence screen first, but keep a clear, defined exit to a full RFP if the screen fails to converge.

The executive's condition, entered directly into the Decision Room, was this: first establish the global common baseline, the regional exceptions, and explicit stop conditions; if no convergence is reached within the deadline, switch to a full RFP. Minerva recorded this response as a receipt and confirmed that the original recommendation held under this added constraint. This is a replay of the executive's actual input and the system's response to it, not a new model conclusion.

A full RFP can formally capture local procurement and compliance detail, but it is slow, expensive, and does not guarantee that regional definitional or political conflicts get resolved. The time-boxed fast evidence screen can surface a common baseline and necessary regional differences at far lower cost, with the tradeoff that an unclear scope could still miss local conditions. The executive chose the second path, precisely because it was stoppable and reversible rather than locking in months of RFP spend before the baseline question was even answered.

The committed next action is to define the fast screen's time-box, name an accountable owner, set explicit comparison criteria, and write down the stopping and escalation conditions before the screen launches. In the underlying public case, the vendor selection team completed a comparable fast evidence process in roughly three weeks before reporting back for sign-off. Minerva does not select a vendor on the company's behalf and makes no claim about return on investment.

This run passed all ten out of ten decision-quality checks. It used four model calls, delivered its first decision in 17.114 seconds, and completed the full result, including the three-advisor cross-check, in 26.135 seconds, passing both the thirty-second first-decision threshold and the forty-five-second complete-result threshold. This remains a single case test of the product's evaluation process. It demonstrates decision-support performance only, not production-scale service performance or real customer outcomes.