AI Meeting Notes for Skip-Level 1:1s
A single skip-level conversation is one data point, not a verdict on a team. The value only shows up once several of them are compared, and the whole practice depends on nobody suspecting the notes traveled back to their manager.
One skip-level tells you almost nothing on its own
A skip-level 1:1 puts a senior leader in a room with someone two or more levels below them, without the person's direct manager present. The pitch is usually framed as access: leadership gets an unfiltered read on the team, and the employee gets a channel to someone who can actually act on structural problems their manager cannot fix alone.
The problem is that one conversation is a sample size of one. The person might be having their best month or their worst week. They might be still annoyed about a decision from three days ago, or newly excited about a project that has nothing to do with how the team is actually doing. A leader who walks out of a single skip-level and treats it as representative of the team's state is making a call on noise, not signal.
The real value shows up only once a leader has done several of these across a team and can compare them. If three unrelated people, in three separate conversations weeks apart, independently raise the same friction point without being prompted, that is a finding. If one person raises it once, it is a data point that might matter or might not. Nothing about a single meeting can tell the difference. Only a record that survives past the meeting, and can be set next to other records, can.
Why the practice exists at all
Skip-levels are a structural fix for a structural problem: information degrades on the way up. A direct manager filters what they escalate, sometimes to protect their team, sometimes to protect themselves, and usually without meaning any harm by it. By the time a concern reaches a VP through the normal chain, it has often been softened, reframed, or dropped.
The Gallup research cited by Lattice puts a number on the underlying trust gap this is meant to close: only 21 percent of employees strongly agree that they trust their organization's leadership. A skip-level is one of the few mechanisms that lets a senior leader hear something closer to the unedited version, directly, without asking a manager to relay it. Lattice's own survey work adds a retention angle: 76 percent of employees said they would likely leave a role if growth opportunities felt limited, which is exactly the kind of concern a direct manager may not be positioned to raise on someone's behalf.
None of that works if the leader treats the conversation as a one-off. The entire premise, hearing something a manager would not have surfaced, only pays off if what gets heard is retained, compared against what other people said, and acted on. A skip-level that is not followed by anything is a listening exercise with no output.
The confidentiality problem is the actual design constraint
The single fastest way to kill a skip-level program is for someone to suspect that what they said made its way back to their manager, even in a softened or paraphrased form. It does not need to actually happen. A manager mentioning something in a later 1:1 that lines up too closely with what a report said privately in a skip-level is enough to end candor for that person, and often for everyone who hears about it afterward.
This is why the guidance from both 15Five and Lattice converges on the same instruction from different directions. 15Five's preparation guidance for the employee side warns against making the skip-level about criticizing the direct manager in detail, precisely because specific complaints are the easiest thing to trace back to a source. Lattice's guidance for the leader side is more direct: keep confidentiality by noting down key themes or topics rather than a blow by blow of who said what, and use role-based visibility so that raw notes are not something a direct manager, or anyone outside the leader running the skip-level, can casually access.
A written record raises the stakes on this in both directions. It makes the aggregate pattern across many conversations visible for the first time, which is the entire point. It also makes a specific, timestamped, attributable statement easier to leak or misuse than a half-remembered conversation ever was. The record is only worth building if the access boundary around it is at least as strong as the boundary a leader would enforce by memory alone.
The other failure mode: nothing happens afterward
HR consultant Joey Price's warning, cited by Lattice, describes the second way skip-levels fail, and it is the opposite problem from a leak: silence. "When an employee is left to wonder whatever happened as a result of their meeting with you, they rarely fill their minds with positive thoughts." An employee who raised something real and then heard nothing back concludes the meeting was theater, and they are usually right to conclude that, because a leader without a record has no reliable way to remember what they promised to look into three weeks and four other skip-levels later.
15Five's guidance names the same failure from the other seat: failing to follow through on action items after a skip-level damages credibility on both sides, and it recommends employees send a short note afterward and keep their own record for the next meeting. That employees feel they need to keep their own paper trail, in case the leader does not, is itself a sign of how often this step gets skipped.
Lattice's recommendation is to follow up with the employee directly within roughly a week, close to the meeting, before the specifics have gone soft in memory. That timeline only holds if there is something concrete to follow up on, which requires the conversation to have been captured well enough to act on later instead of relying on a general impression.
What a skip-level record should actually contain
The two failure modes above point at what the record needs to do and what it needs to avoid doing. It needs to preserve enough specificity that a theme raised independently by three different people, weeks apart, is recognizable as the same theme rather than three vague impressions. It needs to preserve enough detail that a leader can follow up on a specific commitment rather than a general sense that something was discussed.
At the same time, it needs strict per-conversation access control, not organization-wide searchability. A skip-level record is not a QBR or a roadmap review, where broad visibility is the point. The value depends on exactly the opposite: a small number of people, typically just the leader who ran the conversation, being able to see the raw content, while an aggregate view, themes and patterns without attribution to a specific person or quote, is what gets used to spot what several people are independently saying.
Botless capture matters here for a narrower reason than in most meeting types: skip-levels are already sensitive enough that adding a visible bot or a separate recording participant to the call changes what people are willing to say, on top of whatever hesitation already exists about talking to someone two levels up. A record that comes from the conversation itself, without a third party sitting in the call, removes one more reason for someone to hold back.
A practical shape for running these
Run skip-levels on a fixed cadence with each person, not as one-off events triggered by a concern. A single meeting that only happens when something is already wrong will read as an investigation, not a listening channel, and it will get investigation-quality answers.
Tell the direct manager the skip-level is happening, without disclosing content. 15Five's guidance recommends proactively updating the direct manager rather than treating the meeting as something to hide from them; the confidentiality boundary is around what was said, not around the fact that a conversation took place.
Review the record across a team periodically, not after every individual conversation. Patterns that would be invisible from any single skip-level, three people independently naming the same blocker, or the same praise for a specific decision, become visible once several conversations sit next to each other. That comparison is the actual deliverable of the practice, and it is the reason a written record earns its keep here more than in almost any other recurring meeting.
Sources
FAQ
How is this different from a regular 1:1 with a direct report?
A regular 1:1 is a standing conversation with someone's own manager, focused on their day to day work, feedback, and growth plan, and the manager is expected to act on it directly. A skip-level involves someone two or more levels below the leader, happens far less often, and its value depends on comparing it against other skip-level conversations rather than acting on any single one in isolation.
Should the notes from a skip-level be shared with the person's direct manager?
Not the specific content. Sharing what was said, even paraphrased, risks the report identifying themselves as the source and shutting down candor in every future skip-level. If something needs to change, it should usually be acted on by the leader who heard it, or raised as a general theme without attribution, not passed down as a direct quote.
How often should skip-levels happen?
Often enough that a leader can compare multiple conversations with the same people over time, commonly quarterly or twice a year per person for a large team. A cadence that only produces one data point per person per year makes the comparison this practice depends on much harder to do.
What should actually get tracked after a skip-level?
Any specific commitment the leader made, so it can be followed up on within roughly a week while it is still fresh, and the general theme of what was raised, kept separate from a verbatim transcript that could be traced back to the person who said it.