Artificial intelligence adoption is no longer just a technology strategy. In many workplaces, it is becoming a performance metric. According to a new report from the Financial Times, some employees at Amazon are increasing their use of internal AI tools — sometimes by automating unnecessary tasks — to appear more engaged with the company’s AI push.
The behavior has reportedly become common enough to earn its own nickname inside tech circles: “tokenmaxxing.”
The trend offers an early glimpse into a new workplace reality where workers are not only judged on productivity, but also on how visibly they use AI.
What Is Happening Inside Amazon?
The report says Amazon has expanded the use of an internal AI platform called MeshClaw, a system designed to help employees create AI agents that can independently perform tasks.
These AI agents reportedly can:
- Connect with workplace software
- Assist with coding
- Draft or manage emails
- Interact with platforms like Slack
- Automate repetitive workflows
The company has also introduced AI adoption targets, encouraging employees to use AI tools regularly.
One of the key metrics being tracked is token consumption.
What are AI tokens?
In generative AI systems, tokens are units of data processed during prompts, responses, and other AI operations.
A token may represent:
- A word
- Part of a word
- A symbol
- A fragment of code
The more prompts and tasks an employee runs through an AI model, the higher their token usage becomes.
That measurement is now reportedly being used internally as one signal of AI engagement.
A useful infographic here could explain:
- What AI tokens are
- How token usage is measured
- Why companies track AI activity
What Is ‘Tokenmaxxing’?
“Tokenmaxxing” refers to employees deliberately increasing AI activity — sometimes beyond what is practically necessary — in order to appear more aligned with company expectations around AI adoption.
The practice reportedly includes:
- Running simple tasks through AI unnecessarily
- Automating low-value workflows
- Generating extra prompts to increase usage metrics
- Creating AI-assisted processes mainly to show engagement
The term mirrors internet slang where “maxxing” refers to optimising a specific metric or trait as aggressively as possible.
In this case, the metric is AI usage itself.
Why workers may feel pressured
As companies race to integrate AI into everyday operations, employees increasingly fear being seen as resistant to change if they do not actively use these tools.
That pressure can create a workplace incentive structure where:
- Visible AI usage matters as much as actual productivity
- Employees compete to appear technologically adaptable
- Metrics become performative rather than meaningful
This is not unique to AI.
Corporate history is full of examples where workers optimized around measurable activity instead of genuine outcomes:
- Sending late-night emails to signal dedication
- Scheduling excessive meetings
- Inflating productivity dashboards
- Prioritising measurable tasks over valuable work
AI adoption metrics may simply be the latest version of that phenomenon.
Why Companies Are Pushing AI Adoption So Aggressively
Tech companies are under intense pressure to prove they are integrating AI across their businesses.
Executives increasingly view AI as critical for:
- Reducing operational costs
- Increasing productivity
- Accelerating software development
- Automating internal processes
- Staying competitive
For companies like Amazon, widespread internal AI adoption also serves another purpose: training employees to work alongside AI systems before those tools become deeply embedded across every department.
AI literacy is becoming a workplace expectation
Many firms now treat AI familiarity similarly to how they once treated:
- Spreadsheet skills
- Email proficiency
- Cloud software knowledge
In other words, AI usage is quickly shifting from optional to expected.
That shift helps explain why some employees may feel compelled to demonstrate AI activity even when it adds limited value.
The Risk of Measuring the Wrong Thing
The Amazon report raises a larger question confronting many organisations: how should AI adoption actually be measured?
Tracking token usage may offer a simple quantitative metric, but it does not necessarily reveal:
- Whether work quality improved
- If productivity increased
- Whether employees saved time
- If AI outputs were accurate
- Whether automation created business value
Metrics can distort behaviour
Management experts have long warned about what happens when organisations rely too heavily on measurable activity.
A famous principle known as Goodhart’s Law states:
“When a measure becomes a target, it ceases to be a good measure.”
If employees are rewarded for AI usage itself, they may naturally optimise for usage volume rather than meaningful results.
That can create inefficiencies, including:
- Wasteful AI queries
- Increased computing costs
- Redundant workflows
- Lower-quality outputs
- Overreliance on automation
Ironically, companies trying to improve productivity through AI could end up encouraging performative AI behavior instead.
Workers Are Excited About AI — and Nervous Too
The push toward AI-heavy workplaces is happening alongside growing worker anxiety about automation.
Employees across industries worry about:
- Job displacement
- Increased surveillance
- Performance monitoring
- Skill obsolescence
- Pressure to constantly adapt
Women may face disproportionate disruption
Recent research has suggested that women workers could face higher exposure to AI-driven job disruption because they are heavily represented in administrative and support roles vulnerable to automation.
Some studies estimate that women account for a large majority of workers in positions most susceptible to AI replacement or restructuring.
That concern has fueled broader debates around:
- Bias in AI systems
- Workforce retraining
- Economic inequality
- Responsible automation policies
Why the Amazon Story Matters Beyond Amazon
The bigger significance of the report is that it may preview the next stage of workplace AI culture.
The first phase of generative AI adoption focused on experimentation:
- Trying chatbots
- Testing coding assistants
- Exploring automation tools
The second phase may be about measurement and accountability.
Companies increasingly want proof that employees are:
- Using AI tools
- Adapting workflows
- Improving efficiency
- Integrating automation into daily routines
That creates a new tension:
How do organizations encourage innovation without incentivising meaningless AI activity?
The answer could shape workplace culture across industries over the next several years.
AI Metrics Could Become the New Productivity Score
Many businesses already track:
- Emails sent
- Sales calls completed
- Lines of code written
- Customer tickets resolved
AI usage metrics may soon join that list.
But unlike traditional productivity measurements, AI activity introduces new complexities:
- More usage does not always mean better work
- Automation quality can vary dramatically
- Employees may use AI differently depending on their role
- Excessive reliance on AI can create compliance and accuracy risks
For managers, the challenge will be distinguishing between:
- Valuable AI integration
- Cosmetic AI engagement
That distinction may become one of the defining workplace management problems of the AI era.
