AI Features FAQ

This document answers frequently asked questions about AI-powered features in Trakstar Perform, including how they work, how your data is handled, and what safeguards are in place.

What AI features are available in Trakstar Perform?

Trakstar Perform includes an AI-powered Notes Summarization feature. During the review scoring process, managers and/or employees (based on your organization's permission settings) can click a "Summarize" button to generate a concise summary of the notes collected on a given competency or goal throughout the review cycle. The summary appears in-line and is intended as a drafting aid to help users write more thorough, well-informed review sections.

Perform’s AI Drafting Assistance helps managers and employees improve feedback they've already written inside a Perform text field. The user writes a comment, note, or answer. They select Refine and pick an action: fix spelling and grammar, improve the writing, change the tone, shorten it, or lengthen it. Perform AI shows the reworked version in a window directly below the field, where the user can accept it, refine it further, copy it, ask for a different attempt, or throw it away.

How does the AI generate summaries?

When a user clicks the Summarize button, the notes associated with that competency or goal are sent to the AI model, which produces a short summary synthesizing the key themes and observations from those notes. The summary is returned in-line for the user to review, edit, or discard before incorporating it into their review.

Refine works from the text the user has already written in a specific field — the user picks one of the fixed menu actions (fix spelling and grammar, improve writing, change tone, shorten, or lengthen), and Perform’s AI model returns a reworked version in a window below the field. The model is constrained at the system level to rework only what the author wrote: it cannot introduce metrics, dates, names, incidents, ratings, or accomplishments that aren't already there, and each action carries its own additional constraints. It has no access to the employee's notes, goals, prior reviews, ratings, or 360 feedback, and there's no free-text prompt — so the only inputs are the user's own words and the action they selected.

Does the AI make recommendations about employment decisions?

No. The AI is explicitly designed to summarize the content of existing notes only. It will not generate recommendations or language related to promotions, hiring decisions, terminations, bonus eligibility, or performance improvement plans (PIPs). This restriction is enforced at the system level through prompt-level guardrails, and is continuously validated through our reliability testing framework.

Is there a disclaimer when AI-generated content is displayed?

Yes. Every time a summary or refinement is generated, a disclaimer is displayed alongside the output. The disclaimer reads:

• “AI summaries do not reflect performance. Review for accuracy and use your judgment before taking action.”
• “AI-generated content may not fully reflect your intent. Please review carefully and use your judgment before proceeding.”

The disclaimer is persistent and visible without requiring any additional clicks or interactions. It appears each time a summary or refinement is generated.

Can the AI summary affect or override a performance rating?

No. The AI summary is a drafting aid only and has no connection to the scoring or rating system. It does not assign, influence, or reflect any official performance rating. The reviewer retains full control over all scores and written content in the review.

How is my organization's data handled?

Employee and review data processed by the AI feature remains within the Trakstar Perform platform. The AI model's privacy practices and security operations have been reviewed and approved by Mitratech's SecOps, Risk, and Compliance teams. Personal data is kept confidential and is not shared with third parties as part of the summarization process.

Does Trakstar keep records of AI-generated summaries?

Yes. Trakstar maintains a full audit trail of AI activity, including the inputs provided to the model and the summaries it generates. These records are available in the event that an organization needs to review how a particular summary was produced — for example, if there is a question about the basis for content included in a performance review.

Is the AI feature optional?

Yes. AI features in Trakstar Perform are controlled by your organization's super admin. They can be enabled or disabled at any time through admin settings, and access can be configured separately for managers and employees. The feature is off by default and must be explicitly turned on before any users can access it. No user is ever required to use the Summarize button — it is an optional tool available during the review scoring process.

How does Trakstar address bias and fairness in AI-generated content?

The AI model is instructed to use neutral, professional language appropriate for a performance review context. Trakstar actively tests and refines its AI prompts to minimize the risk of biased, editorialized, or inappropriate language in summaries. Summaries are evaluated against a reliability testing framework that includes tone, accuracy, and safety as scored dimensions.

Who can I contact with additional questions?

If you have further questions about Trakstar Perform's AI features, data handling practices, or security posture, please contact your Trakstar account representative or reach out to our support team at perform@trakstar.com.