Fabio the Robot Was Built to Help Shoppers. It Drove Them Off Instead.
What happened
A robot that can't answer a shopper's question sounds like a punchline, not a case study. But when that robot is standing in the aisle of a real store, greeting real customers who came in to buy groceries, the joke points to something more serious: nobody checked whether it was ready before it met the public.
Researchers at Heriot-Watt University in Scotland built the robot, named Fabio, and placed it inside Margiotta, a grocery store, to see how it handled the job of a floor assistant. The idea was simple. Shoppers would ask Fabio where to find something or what to buy, and the robot would point them in the right direction. In practice, its answers missed the mark often enough that Margiotta reported customers avoiding it altogether. A robot meant to draw people in ended up pushing them toward the door.
The failure here isn't really about a clumsy chatbot in a plastic shell. It's about the gap between a lab demonstration and a live commercial floor. A prototype that struggles with edge cases in a controlled trial is a research finding. The same prototype struggling in front of paying customers, with no one tracking how often it got things wrong or deciding when performance was bad enough to pull it, is a business risk dressed up as an experiment. Somewhere between the university lab and the shop floor, the project skipped the step where someone sets a bar for acceptable performance and watches whether the system clears it.
That gap is the real story. Nobody seems to have owned the decision about when Fabio's answers were good enough to keep it in front of customers, and nobody has a documented record of how many interactions it got wrong before Margiotta noticed the pattern on its own.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.