Betterworks helps organizations turn performance into measurable business impact. Built for large and fast-growing companies, its AI-native solution brings goals, feedback, skills, and talent intelligence into the flow of work — giving leaders real-time visibility into workforce capability, manager effectiveness, and progress against business priorities. Betterworks helps organizations align teams, coach employees, support growth and internal mobility, and make confident talent decisions grounded in evidence rather than guesswork.
Managers are already using AI to prepare for performance reviews, summarizing feedback, suggesting goals and drafting language for conversations with their teams.
We can debate whether they should be doing that. But that debate is quickly becoming irrelevant. The reality is they are doing it.
A 2026 survey commissioned by Okta found that more than half of knowledge workers had used AI tools their employer hadn’t approved, with many acknowledging that they’d shared confidential business or HR information with those tools. At the same time, most executives surveyed believed they had visibility into how AI was being used.
That disconnect should get the attention of HR and IT leaders. Not because AI is inherently the problem. Because we need to pay much more attention to the information the technology is working from.
AI can make an incomplete picture sound complete
Think about what happens when a manager asks an AI assistant to help write a performance review.
If the manager prompts it with a few notes, emails and messages, the AI will produce an answer based on those inputs. It may be articulate and well-organized. It may even sound objective.
But it still only knows what it was given.
Recency bias, visibility bias and other inconsistencies in human judgment don’t disappear when AI enters the process. If anything, we have to be more conscious of them because AI can give incomplete information a veneer of authority.
I’ve led and scaled large teams for much of my career. And when managers struggle, it’s because they’re busy, they have imperfect memories, and the information they need is too scattered.
Performance management, for example, has operated this way for decades. AI changes the scale of the problem. A judgment based on incomplete information can now be generated in seconds and delivered with higher confidence.
Employment lawyers are already warning about this issue. When the underlying information is incomplete or skewed, AI can reproduce or amplify existing gaps.
The rub: AI doesn’t magically correct the information we give it.
The problem is the context.
Before asking what an AI system can generate, organizations need to ask what information it can actually access.
A manager preparing for a performance conversation shouldn’t have to reconstruct an employee’s year from memory. There may already be a much richer record: goals established months earlier, progress against those goals, feedback received over time, recognition and previous conversations.
If the AI can’t securely reach that information, the manager is copying-and-pasting. That creates two problems at once: incomplete context and unnecessary risk around sensitive employee information.
This is where HR and IT need to work together. HR understands the performance context managers need. IT grasps why access, permissions and governance matter. AI makes those two conversations inseparable.
Secure access is becoming part of good performance management
New approaches such as Model Context Protocol, or MCP, create a standardized way for AI assistants to connect with other systems instead of requiring people to manually move information between them.
MCP enables approved AI assistants to access performance context while carrying through the permissions that determine what each user is allowed to see.
The larger idea matters more than the protocol itself.
Instead of giving an AI assistant whatever information a manager happens to remember, organizations can give approved tools access to the performance context that manager is already authorized to see.
More complete context doesn’t make human judgment perfect. Ultimately, managers still need to question the output, apply judgment and own the decision. But giving people better evidence to work from can help make those decisions more informed and consistent.
That’s a much more useful role for AI in performance management than writing a punchier paragraph.
Ask what the AI can see before asking what it can do
What would I do if I were evaluating an AI tool that could influence performance conversations or talent decisions? I’d ask two questions first:
What information can this system see?
Why is it allowed to see this information?
If the first answer is “whatever the manager pasted into it,” you have an information problem. If nobody can clearly answer the second, you have a governance problem.
AI isn’t going to remove human judgment from performance management. Rather, AI’s real value is helping people make better use of information that already exists without spending hours finding and assembling it first.
We’ve spent years asking whether AI can write a performance review.
The question that really matters is whether it can help a manager see enough of the work to make a better judgment in the first place.
Learn more about Betterworks at https://www.betterworks.com/
Related News:
Prophecy Launches AI Data Prep Tools and Expands Agents to Snowflake
