GitHub CEO Thomas Dohmke has publicly endorsed a controversial Microsoft memo that ties employee AI usage to performance evaluations, calling it “totally fair game” — and just four days later announced his departure from the company. The dual whiplash has reignited debate over how tech giants should navigate AI adoption mandates, measurement, and workplace culture.

Appearing on the Decoder podcast on August 7, Dohmke told host Nilay Patel that asking employees to reflect on their use of tools like GitHub Copilot, Microsoft Copilot, and Teams Copilot is consistent with a growth-oriented mindset. “I think in 2025, it’s totally fair game to say you should reflect on your AI usage, and you should reflect what did you learn about AI,” he said. The comments were a full-throated defense of an internal memo first reported by Business Insider in June, which had drawn sharp criticism from employees and industry observers.

The memo, authored by Julia Liuson, president of Microsoft’s developer tools organization, instructed managers to consider AI tool usage as part of “holistic reflections” on employee performance. It declared that “AI is now a fundamental part of how we work” and that using AI is “no longer optional — it’s core to every role and every level.” Some teams, according to multiple reports, had already begun weighing formal metrics tied to AI adoption for upcoming review cycles.

Dohmke argued that the memo was “more nuanced” than headlines suggested and centered on manager-employee conversations about learning rather than blunt productivity quotas. He acknowledged that simplistic measurements like counting lines of AI-generated code would be “easily gamified,” but stood by the principle of evaluating an employee’s engagement with internal tools. “It’s about demonstrating a mindset that aligns with company culture,” he said.

That culture, he added, already demands that every GitHub employee — from engineers to HR to legal — use the GitHub platform. “There is no world where I would allow for somebody to say, ‘Well, sorry, I don’t want to use GitHub.’ And I think that’s fair game,” Dohmke said, adding that employees who disagree could seek opportunities elsewhere among “tens of thousands of other tech companies.”

The memo lands at a precarious moment for Microsoft, which has invested billions in Copilot-branded AI assistants but faces internal adoption gaps and external competition. Executives overseeing products like GitHub Copilot have a clear incentive to ensure their own teams use and improve the tools. Liuson’s memo can be read as a cultural nudge toward that goal — but it also risks being interpreted as a mandate that could erode trust and inspire metric-gaming if not carefully implemented.

Employee Backlash and the Perils of Measuring AI Adoption

News of the memo sparked immediate unease among staff and broader industry commentary. For many employees, the directive felt like a shift from optional productivity enhancement to mandatory behavioral compliance. “Forcing tool usage is different from training and convincing employees of value,” noted one analyst, capturing a sentiment echoed across forums and internal channels.

The risk of turning AI adoption into a checkbox exercise is real. Without clear guardrails, managers under performance pressures may prioritize raw usage statistics — such as Copilot suggestions accepted or meetings summarized — over meaningful integration. Such metrics are brittle: they can be gamed, they reward output over quality, and they ignore legitimate reasons for non-use, including security, privacy, and accessibility constraints.

Dohmke himself warned against simple quantitative measures, but the damage to morale and trust begins as soon as employees suspect they are being watched for compliance rather than coached for growth. As one anonymous Microsoft employee told a tech publication, “If my manager starts counting how many times I use Copilot, I’ll just open it in the background and ignore it — but I’ll still get the credit.”

Mandating AI tool usage treads into legal and regulatory gray zones. In jurisdictions with strict employment laws, tying AI use to compensation or career progression could face challenges if the measurement methods produce disparate impacts. Moreover, many AI assistants log prompts, code context, and meeting transcripts to improve models. When those logs contain proprietary, regulated, or personal data, mandating usage without rigorous data governance can create serious compliance exposures.

Large organizations typically require security reviews before deploying new tools, but Microsoft’s push to increase adoption may skip or accelerate those workflows. For teams handling sensitive information — legal, HR, finance — the mandate could inadvertently cause data leakage into unsecured logging pipelines. Dohmke’s insistence that all roles must use GitHub, for example, does not automatically address the data stewardship concerns of a salesperson discussing confidential client strategy with a Copilot tool.

What Managers Should Actually Measure (and Avoid)

For companies considering a similar approach, the memo offers a case study in what not to do. Any performance evaluation tied to AI usage must be grounded in clear principles: purpose, privacy, transparency, fairness, and development. Managers should distinguish high-signal indicators from noise:

  • Meaningful signals: completed AI training, documented use cases, sharing best practices, thoughtful integration (e.g., using Copilot for prototyping with rigorous code review afterward).
  • Surface-level noise: raw counts of accepted suggestions, lines of auto-generated code, frequency of opening an AI pane without evidence of value.

Operational checklists for HR should include publishing explicit rubrics, training managers to coach rather than penalize, maintaining documented opt-out processes, and conducting regular audits to ensure metrics aren’t gamed or disadvantaging certain groups.

The Deeper Cultural Signal — and Dohmke’s Departure

Dohmke’s defense was more than a podcast soundbite; it was a cultural signal that GitHub, even within Microsoft’s sprawling org chart, expects dogfooding as a non-negotiable norm. That expectation can accelerate product improvement and build internal empathy, but it also amplifies the pressure on managers to enforce it sensitively.

Then, on August 11, Dohmke announced he would step down as CEO at the end of the year, writing in a blog post that GitHub’s leadership would remain part of Microsoft’s CoreAI organization. The timing — four days after his Decoder appearance — raised eyebrows across the industry. While no direct connection was drawn to the memo controversy, leadership transitions often recalibrate internal priorities. The next head of GitHub may interpret Liuson’s memo differently, or double down on its principles.

For now, the episode crystallizes a modern dilemma: companies must learn and iterate with AI to stay competitive, but pushing adoption through performance pressures risks corroding trust, creating perverse incentives, and exposing legal vulnerabilities. The path forward lies in aligning incentives with learning and quality, not raw usage counts — and recognizing that mandate and motivation are not the same thing.

Microsoft has not released official HR guidance or concrete rubrics, and many details remain opaque. Developers, managers, and industry watchers will be looking for signs of whether the memo’s intent translates into a thoughtful, empathy-driven framework or a blunt instrument of compliance. In a talent market where autonomy is prized, the answer could shape recruiting, retention, and the culture of AI-first workplaces for years to come.