Meetings & People
How to Write Performance Reviews Using AI Notes (Without Recency Bias)
Write performance reviews grounded in a full year of evidence, not the last three weeks. AI notes eliminate recency bias by synthesizing months of 1:1s.
Review season is approaching. You have six direct reports. HR wants calibrated ratings, specific examples, and development plans for each person. You open a blank document and try to remember what happened over the past six months.
The first two names are easy. You have strong impressions, recent interactions, clear examples. By the third name, things get blurry. You know this person has been solid, but you struggle to point to specific moments. By the fifth name, you are pulling examples from the last month because everything before that has faded into a general feeling.
This is recency bias, and it is the most pervasive distortion in performance management. Research consistently shows that managers overweight recent events when evaluating performance over longer periods. The person who had a rough start but finished strong gets rated higher than the person who was consistently excellent but whose recent work was unremarkable. The person who made a visible mistake last month gets a harsher review than their full-year performance warrants.
The problem is not intention. Most managers want to be fair. The problem is that human memory is structurally incapable of holding six months of observations across multiple people with equal clarity. You need an external system that captures evidence throughout the year and makes all of it accessible at review time.
The Year-Long Capture Habit
Fair performance reviews start months before review season. They start with the habit of capturing observations as they happen -- not in a performance management tool you log into quarterly, but in the same place where you already take meeting notes.
The pattern is simple. Every 1:1 produces a note. Every notable observation -- positive or negative -- gets captured as a quick note. Every coaching conversation, every piece of peer feedback, every project milestone that reveals something about how someone works. These notes go into a collection for each person you manage.
What to capture in a 1:1 note:
What you discussed -- the actual topics, not just "status update"
What they said about their own work -- their self-assessment in the moment is often more honest than the one they write months later
What you observed about their approach -- not just the outcome, but how they got there
Action items -- yours and theirs
Personal context -- the life circumstances that affect their work, shared in trust
A single 1:1 note takes two to three minutes to capture. Over a year of weekly meetings, that accumulates into fifty notes per person -- a detailed longitudinal record of their performance, development, and the relationship between you.
For managers with large teams, the collection-per-person pattern scales. Whether you have five reports or forty, the workflow is the same: capture after each interaction, query when you need synthesis. The AI handles the volume. For the full 1:1 workflow, see our guide on running team meetings from your notes app.
The Review Query That Changes Everything
When review time arrives, open Mem Chat and ask a question you could never answer from memory alone:
"Based on all my notes this year, what are the key themes in this person's performance? Include both strengths and areas for development, with specific examples from throughout the review period."
Chat reads across every 1:1 note, every observation, every project discussion in that person's collection. It produces a synthesis that spans the full year -- not weighted toward recent events, but drawing from the complete record.
The output typically reveals things you had forgotten. A coaching moment from Q1 that planted a seed visible in Q3 results. A struggle in Q2 that the person worked through and resolved by Q4. A consistent strength that shows up in note after note but that you had stopped noticing because it became the baseline.
This is not the AI writing the review. It is the AI showing you the evidence. You bring the judgment -- the calibration, the context the notes cannot capture, the understanding of organizational expectations. But the evidence base is comprehensive instead of skewed.
Other queries that managers find useful during review prep:
"What commitments did I make to this person about their development, and which ones have been fulfilled?"
This surfaces your side of the equation. Reviews are not just about the employee's performance. They are about whether you held up your end of the development relationship.
"What did this person say about their career goals across our 1:1s this year?"
Their aspirations, captured in their own words over multiple conversations, inform a development plan that actually reflects what they want -- not what you assume they want based on the most recent discussion.
"How has this person's approach to [specific skill] changed over the past six months?"
This is the growth query. It traces development over time, making progress visible that is invisible in any single meeting. For more on tracking development arcs, see our guide on tracking leadership growth with AI notes.
Writing Reviews That Hold Up
A performance review grounded in a year of captured evidence looks fundamentally different from one written from memory. Here is the contrast:
From memory: "You've been a strong contributor to the team. Your work on the Q3 project was excellent. I'd like to see you take on more leadership responsibilities next year."
From notes: "Your work this year showed a clear progression in how you handle cross-functional coordination. In Q1, the integration project required significant support from the team lead to navigate stakeholder alignment. By Q3, you were independently running the vendor evaluation process, proactively pulling in the right people and making decisions without escalation. The shift was visible in our 1:1s -- in March you mentioned feeling uncertain about pushing back on timelines, and by September you were the one holding external partners accountable to deadlines."
The second version is more specific, more developmental, and more likely to land as meaningful feedback. It also demonstrates that you have been paying attention all year, not just assembling impressions at the last minute.
This specificity matters even more for development areas. Vague feedback like "I'd like you to communicate more proactively" gives the person nothing to work with. Evidence-based feedback like "There were three instances this year -- the delayed status update in April, the surprise escalation in July, and the scope change in October -- where earlier communication would have given the team more time to adjust" gives them a clear pattern to address.
Calibrating Across Your Team
Reviews do not happen in isolation. You need to calibrate -- to ensure that your rating of one person is consistent with your rating of another doing comparable work. This is where recency bias compounds: if you are evaluating everyone from recent memory, you are comparing recent impressions instead of full-year performance.
With collections for each person, calibration becomes a structured exercise. Ask Chat to summarize the year for each person, then compare the summaries side by side. Look for inconsistencies:
Is someone rated lower because their recent work was less visible, even though their year-round contribution was strong?
Is someone rated higher because a recent success overshadowed earlier struggles?
Are you applying the same standards to similar roles, or does the vividness of certain examples skew your judgment?
Some managers ask Chat directly:
"Compare the performance themes across my notes for [Person A] and [Person B], who are in similar roles."
The comparison highlights differences in contribution, approach, and development that might not be obvious from separate reviews. It does not make the calibration decision for you, but it ensures the decision is informed by the full record.
The Self-Review That Coaches Itself
The capture-and-query approach works for your own performance review too. Founders and senior leaders who capture coaching sessions, strategic decisions, and reflective notes over the course of a year can query their own development at review time.
"What themes keep coming up in my coaching sessions and leadership reflections?"
"What commitments did I make at the start of the year, and which ones did I follow through on?"
"How has my approach to [hiring, delegation, communication] changed based on my notes?"
These queries turn a self-review from a branding exercise into a genuine assessment. The notes hold what you actually said and did, not what you wish you had said and done. The humility of that record -- seeing the commitments you dropped, the lessons you had to learn three times, the growth that happened slowly -- produces a more honest and more useful reflection.
The Development Plan That Sticks
The review itself is a snapshot. The development plan is what matters next. Most development plans are written during review season and forgotten by February. They fail because they are disconnected from the ongoing 1:1 cadence where development actually happens.
When the development plan is captured in Mem -- as a note in the person's collection -- it becomes part of the 1:1 conversation. Before each meeting, Chat surfaces not just open action items but the development goals you agreed on during the review. "Based on our last review and recent 1:1s, how is this person progressing on their development goals?"
The development plan stops being a document that lives in HR's system and starts being a thread that runs through every subsequent conversation. Each 1:1 note either advances the plan or reveals that it needs adjustment. By the next review cycle, you have a continuous record of development -- not a gap between the plan and the assessment.
Get Started
If you are not already capturing 1:1 notes, start this week. After each meeting, spend two minutes capturing the key discussion points, action items, and observations. Add the note to a collection for that person.
When you notice something noteworthy about someone's work -- a standout moment or a concerning pattern -- capture it as a quick note in their collection. These mid-cycle observations are the most valuable raw material for fair reviews.
Before your next review cycle, ask Chat to synthesize the year for each person. Compare the AI's evidence-based summary to your gut impression. Where they diverge is where bias is most likely operating.
After the review, capture the development plan as a note and add it to the person's collection. Reference it in your 1:1 prep queries going forward.
Performance reviews should reflect reality, not memory. The notes you take today are the evidence you will need in six months. Start capturing, and the reviews will take care of themselves.
