Introducing Evidence-based Review Drafts

Review season is one of the most dreaded moments in the HR calendar. Managers and employees open a blank review form and then spend the next hour hunting through months of scattered 1:1 notes, old feedback threads, goal trackers, and status updates. All in hopes of reconstructing a coherent picture of someone’s performance from memory and fragments. The result is usually a review that over-indexes on the last few weeks, undersells the real work, and takes far too long to write. 

Today, we are introducing Evidence-based Review Drafts to solve these challenges. A new capability of the Lattice AI Agent, it turns months of performance signals already in Lattice into a draft, grounded in real evidence, right inside the review. 

The best part? It’s available now to all Lattice Performance customers at no additional cost. 

From Blank Page to Evidence-backed Draft

Here’s the reality: chances are, your employees are already using AI to draft their reviews. They’re copying job descriptions into ChatGPT, pasting in old feedback, and making changes to the output. The problem isn’t that they’re using AI, but that they are doing so outside your systems, without your data, and without any guardrails to secure it.

Evidence-based Review Drafts brings that workflow inside Lattice, grounded in your actual performance insights, with source citations, permission controls, and a human in the loop at every step. 

Here’s how it works: 

When a reviewer opens a Self, Upward, or Downward review and chooses to draft with the Agent, a context modal shows exactly which data sources were scanned, including 1:1s, feedback, goals, updates, growth areas, and prior reviews, before any draft is generated. Then, for each open-text question, the Agent produces a complete draft that includes the specific insights it drew from the cited sources. 

Reviewers can also check, uncheck, and regenerate those sources, and a "Refine with Agent" sidebar accepts natural-language instructions like "focus more on revenue impact" or "make this more concise." Plus, style controls let reviewers set tone, format, and length across the review. Timeframe controls let them expand or narrow the data window if needed.

Early customers completed reviews 47% faster, on average, by starting with a complete, evidence-backed draft rather than a blank page.

‍And because it draws on data already in Lattice, Evidence-based Review Drafts also makes it easier for employees to keep that data rich: better 1:1s, more frequent feedback, and consistent goal-tracking throughout the year all translate directly into better drafts when review season arrives.

Reviews That Reflect the Whole Period

Recency bias is one of the most persistent problems in performance evaluations. When managers write reviews from memory, recent events crowd out earlier ones and employees get evaluated on a narrow slice of their actual performance. 

Evidence-based Review Drafts is designed to change that. The Agent ingests data across the full review cycle timeframe, not just what happened last month. Every draft ties back to a real data point in Lattice, so reviews become a more accurate representation of what actually happened across the period. 

In my 8 years in the technical field, performance reviews have always been a struggle. When you've packed in months of work, it's almost impossible to remember everything you accomplished. This last cycle, Lattice's AI Agent changed that. It pulled together our 1:1 conversations, past updates, and goals, so I could actually see and advocate for all the hard work I'd put in. — Cody Dauksch, Customer Care AI Specialist @ Lattice

Trusted AI, With Humans at the Forefront

We didn't build Evidence-based Review Drafts to write reviews for managers and employees. We built it to give every reviewer a strong starting point they can trust and to free up the human judgment that makes work meaningful. 

That's been the animating idea behind Lattice AI from the start: AI won't replace us, it will augment us. The goal isn't to automate the employee experience, it’s to make it more human. 

Managers deserve more than a blank page and a deadline. Employees deserve to be evaluated on what they actually did, not on what their manager happened to remember. And HR leaders deserve a review season that's faster, fairer, and grounded in data. 

Evidence-based Review Drafts makes that possible: 

• It cites its sources. Every draft is tied to real Lattice data: a specific 1:1 from October, a piece of feedback from Q2. It won't invent performance that didn't happen.
• It is grounded in context. The Agent assigns a confidence score to every answer. If there isn't enough data to generate a quality draft, it flags the question as "Needs more context" rather than generating something generic.
• It respects permissions. The Agent's context is bounded to exactly what the reviewer already has access to. It won't surface notes or feedback the reviewer can't see.
• It keeps humans in control. The Agent can't select ratings or submit a review. Every review goes through the reviewer, who remains fully accountable for accuracy, tone, and final submission.

Available Today for All Lattice Performance Customers

Evidence-based Review Drafts is available now, included with every Lattice Performance subscription at no additional cost. Log in to Lattice to turn it on for your next cycle.

Not a Lattice customer yet? Request a demo or take an interactive product tour to see Evidence-based Review Drafts and the full Lattice People + AI platform in action.