Researchers who followed a luxury resort in Vietnam for three years — interviews, archival records, weeks of shadowing staff — documented a transformation nobody had planned. In 2014, the resort judged its own service through internal standards, direct observation, and managerial experience. By 2017 it had replaced those with algorithmic analysis of scores pulled from TripAdvisor, Booking.com, Facebook, and Twitter. One guest-relations employee began handing departing guests a gift bag after complaints. No service manual told her to; she wanted to head off a bad review. Staff started checking whether they'd been mentioned by name in ratings. Ranking position turned into an operating target.
Room rates roughly tripled. The resort collected luxury-service awards. Management counted the whole thing a success. What had changed was not only how the resort was measured but what it had begun to optimize for.
The same dynamic shows up wherever structured evaluation carries commercial weight. A September 2026 study of experienced professionals on Upwork — people with thirty years of credentials behind them — found that the platform's five-star categories and algorithmic visibility scores had effectively displaced résumés and references. A former consumer-goods executive spent more than a year filling out surveys and building lead lists, work far below his experience, because that was how you accumulated platform reputation. A software engineer took cheap proofreading jobs. The algorithm pushed successful workers toward more of whatever they had already done well, whatever they might have wanted to do next. Forty percent of the sample left rather than perform the work required to cross the visibility threshold.
And those were algorithms scoring for human buyers who still made the final call. When an agent selects on the buyer's behalf, the structured description stops being how you get shortlisted and becomes how you get chosen, usually without the buyer ever seeing what was filtered out.
The SEO era is the obvious precedent, and it understates the pressure. Restaurants that optimized for search began competing on reviewability — photogenic plating, keyword-dense menu copy — rather than on whatever brought regulars back. But that was mostly a change in how a business presented itself outward. The Vietnamese resort went further. It reorganized how employees were evaluated, redirected their attention toward review mentions, and ran management meetings around platform scores. The gift bags and the name-tracking were downstream of an internal restructuring. Agent evaluation pushes harder in the same direction, because a system that compares can only compare what fits its fields, and the incentive is to cultivate the strengths the system can see.
Some of this is already measurable. A controlled study of AI purchasing agents in a simulated marketplace found that rewriting a single product description produced a statistically significant gain in market share in a third of the scenarios tested, with results varying depending on which model was doing the buying. Not a large effect, and not stable. But it means the description is no longer a representation of the offering. It is an input to the selection.
For businesses that have competed on experiential or relational qualities — a therapist's clinical read of a patient, an architect's sense of proportion, a hotel's atmosphere — the open question is whether specification is a new marketing chore or something that works its way back into the offering itself. The resort suggests the second. The scoring system rewarded different strengths, pulled employee attention elsewhere, and redefined what counted as excellence inside the building. Everyone adapted to the new incentive structure well before anyone sat down to decide whether the trade was worth making.