How Meta, JPMorgan, and Others Are Tracking AI Use — And Why It's Getting Complicated

The Corporate Push to Measure AI at Work

Artificial intelligence is no longer just a tool that employees can choose to use or ignore. Across major corporations, using AI at work has quietly shifted from a novelty to an expectation — and measuring that usage has become the next frontier of workplace management. From Wall Street banks to Silicon Valley tech giants, companies are now investing heavily not only in AI software itself, but in the infrastructure needed to track how, when, and how often employees actually use it.

As corporate America pours billions into AI licensing, software agents, and proprietary systems, executives are under growing pressure to justify that spending. The result? A new wave of internal dashboards, adoption metrics, and monitoring tools designed to quantify something that has historically been very difficult to measure: how productively workers are using artificial intelligence in their daily workflows.

Inside the AI Dashboard Revolution

Companies like JPMorgan Chase and Meta have been among the most aggressive in building out internal tools to track AI adoption. These dashboards typically collect data on how frequently individual employees or teams interact with AI tools, which features they use, how much time they spend on AI-assisted tasks, and in some cases, the outputs those interactions generate.

The motivation is straightforward on the surface. Rising AI costs have made return-on-investment a critical concern. Executives want visibility into whether the billions being spent on technology licenses are translating into real productivity gains — or whether employees are simply ignoring the tools altogether. Internal leaderboards and usage rankings have emerged as one way to create accountability and, in some cases, to pressure employees into deeper integration.

But there is another layer to this data collection effort that goes beyond simple productivity monitoring. In a number of companies, employee interactions with AI systems are being captured and stored not just for oversight purposes, but to help train and refine proprietary AI models. This means that the way employees use AI tools today could directly shape the AI systems their companies deploy tomorrow — a dynamic that raises significant questions about data ownership, consent, and the boundaries of workplace surveillance.

Why Measuring AI Usage Is Harder Than It Sounds

Despite the enthusiasm for dashboards and metrics, the reality of measuring meaningful AI adoption is proving far more complicated than executives initially anticipated. Usage volume — how many times an employee opens an AI tool or submits a prompt — is easy to count, but it is a poor proxy for value. An employee who asks an AI assistant ten shallow questions may generate less business value than a colleague who uses it twice for deeply complex research tasks.

There is also the problem of quality versus quantity. Many of the metrics currently being tracked by companies focus on surface-level engagement rather than outcomes. Counting prompts or logins does not tell you whether the AI interaction actually improved a decision, accelerated a project, or enhanced the quality of a deliverable. Building metrics that capture genuine productivity impact requires far more sophisticated measurement frameworks than most companies currently have in place.

Furthermore, different job functions integrate with AI in fundamentally different ways. A software engineer using GitHub Copilot to write code has a very different usage pattern than a marketing manager using a generative AI tool to draft campaign briefs. Applying a single dashboard metric across both roles risks creating misleading comparisons that reward visible, high-frequency use over genuinely impactful application.

Gaming the System: When Metrics Become the Goal

Perhaps the most revealing consequence of AI usage tracking is what happens when employees figure out they are being watched — and ranked. Across industries, workers have begun finding ways to inflate their AI usage numbers to meet internal expectations or climb internal leaderboards, regardless of whether the activity generates real value.

This kind of metric-gaming is a well-documented organizational behavior. It echoes earlier periods in corporate history when employees padded call logs, inflated billable hours, or gamed customer satisfaction surveys. When measurement systems reward the appearance of productivity over genuine output, employees — rationally — optimize for the measurement. In the context of AI adoption, this can mean submitting redundant prompts, running unnecessary queries, or using AI tools for tasks that could be completed faster and better without them.

The irony is significant: in trying to ensure employees are genuinely integrating AI into their work, companies may be incentivizing performative use that undermines the actual value AI is supposed to deliver.

Privacy Concerns and the Pushback from Workers

Alongside the gaming problem, a more fundamental tension is emerging between employers and workers over privacy. As companies collect increasingly granular data on how employees interact with AI — including the content of prompts, the nature of queries, and patterns of use — employees are raising legitimate concerns about surveillance overreach.

Workers in several industries have begun calling for clearer guardrails around what data is collected, how long it is retained, and whether it can be used in performance evaluations or disciplinary processes. Labor advocates have noted that AI usage tracking represents a meaningful expansion of workplace monitoring capabilities, one that could disproportionately affect certain workers and create a chilling effect on how people engage with AI tools.

• Employees may self-censor or avoid using AI tools entirely if they fear their queries will be judged or misinterpreted.
• Workers in regulated industries face additional concerns about the confidentiality of the data they feed into AI systems.
• There is growing uncertainty about whether employee AI interactions can be used as training data without explicit consent.
• Some employees report feeling pressured to use AI in ways that feel performative rather than genuinely helpful.

The Road Ahead: Smarter Measurement, Not Just More Monitoring

The core challenge that Meta, JPMorgan, and their peers now face is not simply a technical one — it is a philosophical one. Measuring AI adoption in a way that is genuinely meaningful requires moving beyond activity metrics toward outcome-based evaluation. Companies that get this right will likely build cultures where AI becomes a genuine force multiplier. Those that get it wrong risk creating surveillance-heavy environments where employees optimize for looking productive rather than being productive.

For AI investment to truly pay off, organizations need measurement frameworks that are transparent, contextually sensitive, and aligned with actual business outcomes. They also need to earn the trust of their workforce — because in the era of AI, the employees who feel trusted and empowered are far more likely to find genuinely innovative ways to use these tools than those who feel monitored and pressured.

The race to track AI use is well underway. But the real competitive advantage will go to the companies that figure out how to measure what actually matters — and how to do it in a way their employees can actually trust.