Communication Skills Self-Evaluation: Evidence Over Memory

Because a wording problem is usually an evidence problem wearing a disguise. You are being asked to assess a year of conversations you kept no record of.

This is not a writing weakness. It is a records problem, and the records were never kept. Nobody finishes a 1:1 and writes down the moment they restated a decision. At the time it was just a sentence in a conversation.

So six months later you reach for "clear and concise" and "a strong collaborator." Your manager has read those exact words on four other forms this week. They carry no information. Your manager can tell, which is why the section feels hollow to write.

Specificity is the whole game here, and specificity is downstream of having a record. Fix the record and the wording problem mostly dissolves.

It is worth naming what the form actually is: one forced attempt per year to answer a question you probably care about the rest of the year too. Am I any good at this? Did the thing I decided to work on in January change anything? Nobody schedules time to find out, so the review form becomes the only occasion, and by then the evidence is twelve months gone. Everything below works the same whether you have a form due or just want an honest read on yourself.

Three parts: which conversation and what was at stake, what you said or did, and what changed. Short enough to paste into a form, specific enough to act on if you never write one.

A usable example is boring on purpose. It reads like a log entry, not like praise:

• Situation — which meeting, project, or 1:1, and what was unresolved
• Action — the words or structure you used: asked X, restated the decision as Y, sent the summary with owners named
• Outcome — what changed: decision locked, risk surfaced, handoff clean

Worked through: in the March launch sync, engineering and marketing disagreed on the freeze date. You restate both constraints out loud. You confirm the freeze as Friday 5pm with Theo owning the checklist, and post that in the channel before anyone leaves. The freeze holds. No same-day reopen.

That is a stronger sentence than "I communicate clearly with stakeholders." It is also the only kind of sentence you can learn anything from. Two of those a month and you can see your own pattern; a year of "clear and concise" tells you nothing. Either way you can only write it if you still have the details, which is what the next three layers are for.

Go looking for facts in the text — questions you asked, decisions you clarified, dates you named. Not for a personality assessment, which a transcript cannot give you.

Before you open a phrase bank, open a record of your talk-heavy work from the review period. Meetings, 1:1s, client calls, the standup you ran. You are mining for four things:

• Questions you asked that moved a decision forward
• Moments you pinned down an owner, a date, or a scope boundary
• Updates you gave that stopped rework before it started
• Threads you left open and had to chase two weeks later

That last one belongs in the weakness section, and it will be the most credible sentence on the form.

Speech-to-text preserves words and sequence well. It is a poor judge of whether you sounded warm, whether a pause was awkward, or what someone meant underneath what they said. Stay on the mirror: what the text shows you said, and what happened next. If you already capture conversations with Pastok, the Activity timeline and search are how you find those four moments. Beats reconstructing a quarter the night before the form is due.

One transcript is a snapshot. A month of them shows habits, which is the only level at which the question "am I getting better" can actually be answered.

Once there are a few weeks of conversations, a second kind of finding shows up. These are the ones worth putting on a form because a careful human reader could verify them in the transcript:

• You summarize well and then end without naming an owner or a date
• In 1:1s you ask about blockers, then never reopen them the following week
• In group meetings you propose a solution before anyone has restated the problem

Pastok's Communication insights are built for exactly this pass — a second read over what was said across a day or a week. Two boundaries keep it usable rather than embarrassing. First, if a pattern would not be visible in the transcript to a careful reader, it does not belong in a performance document. Tone, warmth, and micro-expressions are easy for a model to assert. They are impossible for you to defend in a review meeting.

Second, check the speaker labels before you claim a line as your own. Attribution goes wrong, particularly when three people talk over each other. A clean quote credited to the wrong person is worse than no example at all.

Review season is a deadline. Getting better is a recurring job. Ask the first set once, then turn the one question you keep asking into a standing instruction.

These are the questions that pull evidence out of a month of conversations. They are worth asking whether or not anything is due. Ask them in Chat, verbatim:

• "Which meetings did I end without naming an owner?"
• "Show me every time I restated a decision out loud."
• "What open questions did I promise to revisit in my 1:1s?"
• "Where did I propose a solution before the problem was stated?"
• "What did I commit to in the March launch sync?"
• "Which conversations ended without a date attached to anything?"

Then pick the one aspect you actually want to change next quarter and stop asking by hand. A Custom Skill is a standing instruction you define once in Chat: what to look at, across which conversations, in what output format. You run it when you want the report.

• "From this week's meetings, list every moment I stated a decision with an owner and a date. Output a Markdown checklist. Say nothing if there are none."
• "From my 1:1s this month, extract the open questions I said I would revisit, and whether I did. Output a two-column table."
• "From client calls this quarter, draft three situation-action-outcome bullets I could paste into a self-evaluation."

Output comes back as Markdown or PDF, which is enough to drop straight into the review form. Skills run when you run them. Nothing grades you overnight.

It gives you dated evidence and a repeatable check. It does not give you a verdict on how you come across, and it is overkill if the form is due tonight.

If all you need is a phrase to paste before midnight, use a phrase bank. It is faster. And be clear about what an annual form can tell you: one self-assessment a year, written from memory, is a bad instrument for a question you are asking about a habit. This method pays off on the second cycle, when you have a record and your peers do not.

Three honest edges. Nothing here reads tone, emotion, or pauses — the analysis runs on transcribed text, and any claim about how you sounded would be invented. Attribution errors are real, so verify a line is yours before quoting it. And this is evidence tooling, not coaching. It can show you that you closed four of eleven meetings without an owner. It has no opinion about what to do next.

Capture has one boundary that has nothing to do with the product. Ask before recording a conversation that includes other people, and follow the recording laws where you work. A self-evaluation is not a reason to tape a colleague quietly.

The test of whether this worked is small. Next April, can you write the communication section in twenty minutes from two dated examples you did not have to remember — and can you say whether the one habit you picked this year actually changed?

• How to prepare for a meeting quickly
• AI meeting notes for consultants