Triumphs, Lessons Learned in Using GenAI for Performance Reviews

One area where the use of generative AI (GenAI) has quickly moved from experimentation to growing adoption is in performance management. As more technology vendors make GenAI tools available on their platforms, HR functions are finding the nascent technology can help with several challenges typically encountered during the performance review process.

Here are the lessons learned as HR leaders find GenAI requires effective guardrails, the right choice of platforms and good employee training to maximize its benefits and minimize its risks.

How GenAI Can Aid Performance Management

Consulting firm Accenture is using GenAI to enhance a variety of tasks in its performance management process, according to Michael Benyamin, a managing director who leads the company's HR transformation and delivery practice.

Benyamin said solutions like Microsoft Copilot can help more effectively cascade performance goals by suggesting objectives and key results that fit a specific employee’s contribution to a shared goal.

Like many organizations, Accenture is working to make performance management a continuous event rather than a periodic or one-time occurrence. “We’ve successfully rolled out a ‘fast feedback’ process in Microsoft Teams to our employees worldwide,” Benyamin said. “The solution allows us to deliver feedback in the flow of work and integrates with Microsoft Graph and Workday to prompt employees to provide feedback for recommended co-workers whom they work with regularly.”

After selecting the tone for the feedback and allowing workers to add additional comments, the fast feedback system generates a complete performance summary that can automatically be added to Workday on an employee’s behalf, Benyamin said.

Accenture will soon be rolling out a GenAI-driven performance summary tool to support more efficient and effective calibration discussions, he said.

“The performance summary creates a consolidated view of performance and development needs by evaluating data such as feedback received, recognition and performance against objectives,” he said.

Bryan Ackermann, head of AI strategy and transformation for consulting firm Korn Ferry, said GenAI tools are proving especially helpful to managers in drafting robust performance review summaries, themes and patterns—as well as in creating development and action plans for employees—during the performance management process.

But he says organizations using GenAI with success also create policies and training that stress the technology should only be used to create first drafts and not finished performance reviews.

“In all cases it’s important to consider GenAI outputs as working drafts,” Ackermann said. “That means checking for inaccuracies, especially in performance summaries and action plans, as well as for proper balance and biased language. As with any GenAI-created HR material, we’d never recommend that it stand alone without review by the individual responsible for the particular element of the process.”

GenAI also can help employees draft narratives about team members’ performance. “The technology can help ensure readability, clarity and completeness of feedback,” Ackermann said.

Vendors Add GenAI to Technology Platforms

Performance management technology vendors have been adding GenAI tools to their platforms at a rapid clip as the technology becomes more mature and accepted within organizations. One such vendor is 15Five, which has a new AI-powered assistant designed to aid in multiple aspects of the performance review process.

Jeff Smith, head of product for Raleigh, N.C.-based 15Five, said one of the chief benefits of the GenAI tool is enabling managers to quickly pull together a first draft of performance reviews from a vast range of text, data and other inputs accumulated during review periods.

“Imagine being a manager with 13 direct reports, and when the performance review cycle begins you’re asked to write 13 separate performance reviews starting from a completely blank slate,” Smith said. “That process subjects employees to a myriad of human biases from their manager, including recency bias, gender bias and idiosyncratic rater bias (where feedback is more reflective of managers/raters than employees being reviewed.)”

Time saved by using GenAI to collect and synthesize performance data, as well as to produce working drafts of reviews, leads to more effective and actionable performance evaluations, Smith said. “With that first draft done managers can spend more of their time refining, personalizing and adding specificity to their feedback,” he said.

Improving Manager Performance

Another AI-based tool on 15Five’s platform is designed to help managers with another traditionally challenging part of performance reviews: taking appropriate actions based on the data and observations gathered during the evaluation process.

“Manager effectiveness is the single most critical driver of performance, engagement and retention, and what managers really need is coaching and growth within the flow of work,” Smith said.

The AI-powered assistant gives managers specific tips for conducting more effective one-on-one meetings with employees, can summarize results from previous review sessions, provide insights into a team’s current state of engagement based on survey results, and offer recommendations for where to take action immediately to better support both individuals and teams, Smith said.

The system also can provide nudges to busy managers, such as, “We’ve noticed it’s been over 30 days since a one-on-one session has been logged for you and your direct reports,” he said.

Help for Employees, Too

GenAI also can help employees as they prepare for their aspect of the performance review process, making it easier and more efficient to create lists of accomplishments, to choose the right vocabulary and to hone narratives they’ll use to state their case while being evaluated.

“Not everyone is great at articulating their own accomplishments or being able to write performance feedback about others, including co-workers or those who might report to you,” Benyamin said. “GenAI is great at synthesizing intent and returning back content that is written for a specific purpose like writing a self-evaluation.”

Cautions About Using GenAI

Despite the benefits of using GenAI in performance management, experts say it’s still important to have the proper guardrails in place to guard against risks like created bias or inaccurate reviews.

Ackermann suggests only using proven tools such as OpenAI’s ChatGPT Enterprise, Bing Chat Enterprise or Microsoft 365 Copilot to reduce risks of using GenAI for performance reviews.

“Those and other GenAI tools don’t have the ability or right to train the model on a user’s prompts, or on the data which is pulled for the answers or the answers themselves,” making them the best bets, Ackermann said. “That way users are able to safely work with confidential performance data.”

Benyamin said given the sensitivity of HR data he recommends using private GenAI tools like Microsoft Copilot for performance management. “This gives a company the opportunity to integrate data from source systems like HCM, sales or recognition platforms to deliver contextualized content,” he said.

Benyamin said there always should be ongoing testing of large language models (LLMs) for intended use cases within HR to ensure bias isn’t being introduced.

“We also recommend having a ‘quick feedback screen’ where users can provide feedback about the effectiveness and bias of generated information,” he said.

Full Disclosure: Being Transparent About GenAI Use

Many HR leaders and consultants also agree it’s important to be transparent with employees about when GenAI is used during any part of a performance evaluation process. That provides reassurances, for example, when employees might think AI is being used to make final performance decisions.

“It’s important to re-assert to employees that the author and the organization are ultimately responsible for the content of the performance review, regardless of the method of drafting the content,” Ackermann said. “I’m not sure I’ve seen much mandated disclosure for internal processes such as performance management, but that may evolve as company adoption (of GenAI) increases.”

Benyamin agrees with the need for clear and open communication about the technology’s use. “Transparency is important to build trust in the use of GenAI,” he said. “If done well, there should be no surprises because both employees and managers are using GenAI for parts of the same process.”

Dave Zielinski is principal of Skiwood Communications, a business writing and editing firm in Minneapolis.