According to a report by the Financial Times, Amazon notified employees this week that its internal AI usage tracking service, Kirorank, has been taken offline. The tool scored and ranked users based on their AI activity on Amazon’s Kiro development platform, with the original intention of pushing development teams to meet the company’s goal of requiring more than 80% of employees to use AI tools weekly. Amazon Senior Vice President Dave Treadwell stated that Kirorank’s starting point was “good,” but the end result was that employees inflated their rankings by “falsely increasing token consumption” (i.e., having AI agents perform meaningless tasks), which incurred additional computing costs for the company. He emphasized to employees: “Please do not use AI for the sake of using AI.” Amazon later confirmed in a statement that this “beta dashboard was not an official or approved tool and has now been deprecated.”
This kind of “ranking manipulation” behavior is not unique to Amazon. Meta employees once competed on token consumption in an internal leaderboard called “Claudeonomics,” where approximately 85,000 employees collectively consumed over 60 trillion tokens in 30 days. That leaderboard was also taken down shortly after being exposed. Microsoft also issued an internal memo characterizing AI usage as a “core requirement for every role,” but later clarified under public pressure that no formal evaluation existed. Analyst Gil Luria commented that using consumption as an incentive metric “doesn’t sound healthy” and warned that quantifying AI adoption rates inevitably leads to score-chasing behavior. As major tech giants collectively push capital expenditures toward $700 billion by 2026, the cost of this “usage arms race” is becoming increasingly real.