Focusing on AI Productivity Over Excessive Usage, ETEnterpriseai
The guidance was circulated internally in September and published on Amazon’s website on Tuesday. Amazon has cautioned its software engineers against excessive use of artificial intelligence (AI) tools simply to boost usage numbers, signalling a shift towards measuring the technology’s impact on productivity and business outcomes rather than the volume of AI consumed.
The company has flagged concerns over “tokenmaxxing”, a practice in which employees or coding agents use large volumes of AI tokens, sometimes without adequately considering the usefulness or quality of the output.
Tokens are the units of text and data that AI models process and generate.
In updated guidance for software development engineers, Amazon said measures such as token consumption, AI tool adoption rates and the volume of code written or reviewed can offer an incomplete picture of how effectively teams are using AI.
Relying too heavily on these indicators could increase expenses and add unnecessary complexity without delivering corresponding productivity gains.
The guidance was circulated internally in September and published on Amazon’s website on Tuesday.
The company said AI is expected to play a central role in the future of software engineering, although effective practices for integrating the technology are still evolving.
The warning comes as technology companies reassess how they measure the benefits of AI adoption amid rising infrastructure and model usage costs.
Several companies had previously encouraged employees to increase their use of AI coding assistants, with internal leaderboards tracking token consumption. However, such incentives can encourage employees to maximise usage rather than focus on the quality and business value of their work.
Amazon shut down an internal leaderboard in May after identifying employees who were using AI tools excessively to improve their rankings. The company has not introduced a cap on employee token usage, according to reports.
The broader industry push towards AI adoption has also been accompanied by questions over whether higher spending and increased usage translate into measurable improvements in engineering output.
Amazon has not ruled out tracking token consumption, code generation or AI adoption.
Instead, the company has said these indicators should be assessed alongside other measures and interpreted with sound judgement.
The emphasis, according to the updated guidance, should be on long-term results rather than the volume of resources consumed or the number of AI-generated changes made.
The company has also advised engineers against using AI tools for tasks where they offer little or no benefit.
Recognising situations in which existing AI tools are not useful should be viewed as a valid assessment, rather than a failure to adopt the technology.
- Published On Oct 28, 2026 at 11:27 AM IST
