Use mentor notes to identify unresolved support needs by recording what was discussed, what changed, what remains difficult and what action was agreed. Keep each note dated and linked to the right participant and enrollment. Review repeated concerns across notes without inventing missing context, and distinguish a suggested follow-up from support that has actually happened.

By Sopact Academy · Updated September 12, 2026. All examples are fictional.

This lesson is for mentoring, coaching and program teams that need to use notes across a caseload. Bring a note template and a few de-identified examples. Leave with writing guidance, an extraction rule and a review list that retains the source behind each observation.

Course connection: notes add context between the baseline, mid-program check-in and exit measurement. They can explain a reported experience or an action; they do not by themselves establish why someone left or what caused an outcome.

A four-step notes workflow

1. Use a short template with clear writing guidance.
2. Separate source statements from interpretation.
3. Check structured outputs against the original note.
4. Review repeated issues, action status and gaps in coverage.

The examples below follow a fictional training program. The same record discipline can be useful in other programs, but note content, review frequency and access must fit the service. Do not infer that a missing note means a missing service or that a repeated barrier predicts departure.

You can test the template and extraction rules with a small set of notes before connecting a live workflow. Software can help organize repeated records and prepare source-linked summaries, but an explicit review process is still needed.

Keep the type of evidence visible. A participant’s statement, a staff observation and an analyst’s interpretation are different. “The participant said they felt discouraged” is a reported statement. “Seems discouraged” is an interpretation that needs attribution and context, not a fact an AI should invent. Do not erase a legitimate first-person account of emotion; record what was actually said and why it is relevant.

What can recurring notes add to a participant record?

A recurring note can document the work attempted, an observed result, a participant’s account of a barrier and the next agreed action. Review what your current notes actually contain before designing analysis. Not every note will cover all four, and an unstated item should remain unstated.

Per-person review and caseload review serve different purposes. Staff need the individual history to understand context; coordinators also need to see unresolved actions and repeated themes across authorized records. A summary should help them find the relevant notes, not imply that the source no longer needs checking.

Recurring notes differ from a scheduled pre/post instrument. Their frequency and content may vary with who wrote them and how often people met. A concern mentioned once can still matter. Repeated wording may reflect copied-forward text rather than a continuing issue, so check dates, authorship and whether the person’s situation was updated.

Step 1 — Design a short, useful note

Four headings are a useful starting point: topics, progress, barriers and next actions. Add the session date, author and participant/enrollment reference. Include other fields only where they serve the program’s requirements. Pilot the template with staff rather than promising every note will take less than five minutes.

Paste this into any AI:

Consistent headings help writers and reviewers find information, but do not guarantee correct extraction. An action field may contain several proposals, a refusal or a completed task. Keep those states distinct and preserve the original note when structuring it.

Step 2 — Define what may be extracted

Define the permitted output before applying AI across records. The extraction rule should identify the source note, quotation or location supporting each item, the observation date and anything that needs review. A citation is a way to check a claim, not proof that the claim is correct.

A quoted passage can still be misread, shortened misleadingly or attached to the wrong person. Check whether the passage supports the extracted claim and preserves negation, timing and speaker. The NIST generative AI risk profile identifies confabulation as a risk; source checking and human review remain necessary even when an output sounds confident.

Step 3 — Check the structured note

A fictional note says the session covered blueprint reading, that the participant completed a task without help, and that they reported transport difficulty. It also says the mentor will raise a possible bus pass or schedule adjustment. The extraction should preserve that distinction: proposing support is not confirming approval or delivery.

Check every structured item against the passage beside it. If the author says “third week,” record that as an author-reported duration; verify dated notes before presenting it as an independently calculated recurrence. In a configured Sopact workflow, this structure can be linked to the participant timeline. Review corrections without overwriting the original source.

Step 4 — Review unresolved actions across the caseload

For caseload review, ask which actions remain unresolved, which records need clarification and which participants lack expected notes. Use the expected session schedule to interpret coverage. A person without weekly notes may have completed the program or follow a different schedule; do not automatically label them unmonitored.

• "Show this person's trajectory" — review dated progress and barrier statements without inferring an unstated emotional state.
• "List every blocker that has repeated two or more weeks for the same person" — verify distinct dates and current status, with supporting passages.
• "Which blockers appear across multiple people?" — a repeated theme raises a program-design question; check current sources and context.
• "Who has no note stream at all?" — compare expected and recorded sessions before interpreting a coverage gap.

Progress and barriers can coexist. A person may demonstrate a new skill while reporting difficulty reaching the venue. Discuss the support they want and record whether the action occurs. A pattern across several people suggests a question about program design; it does not establish that one prescribed fix will work for everyone.

Common mistakes

Blurring statements and interpretation. Keep relevant participant accounts in their own words. Attribute staff interpretations and do not let an AI invent a feeling that nobody reported or observed.

Accepting a citation without checking it. Confirm that the passage supports the conclusion, includes necessary context and belongs to the correct record.

Replacing individual context with a summary. Use cross-record views to find patterns, then return to the relevant source and person before choosing support.

Counting mentions as elapsed time. Three mentions are not necessarily three weeks or three separate events. Check dates, duplicates and whether a later note resolves the issue.

Ignoring gaps in expected notes. Compare recorded sessions with the relevant schedule. Investigate gaps without assuming either that everything is fine or that the participant is at risk.

What you have now

You now have a short note structure, rules for distinguishing evidence types and a review list that keeps actions connected to sources. Test the system on ordinary, ambiguous and incomplete notes before applying it more broadly.

The one thing to do this week

Pilot the revised template with five notes this week. Include one with a resolved barrier, one with a proposed action but no owner and one copied from an earlier entry. Ask two staff members to check the structured output independently. Use disagreements to improve the writing guidance and extraction rule.

Who this is for

Coordinators reviewing many participant records, mentors handing over support, and program leads checking whether proposed actions were completed. The method helps bring relevant notes into those decisions.

‍

A worked check: mentions, dates and unresolved status

Suppose a fictional participant has notes dated September 1, 8 and 15. Each mentions transport. The September 15 note says the issue was resolved on September 10. A count of three mentions is correct; a claim that the barrier remains open is not. Report the latest supported status and preserve the earlier history.

If the September 8 text was copied into the September 15 entry without an update, mark the current status uncertain. Ask for clarification rather than presenting three mentions as three independent confirmations.

Quality checks before using the caseload view

1. Check participant and enrollment identity, observation dates and note authors.
2. Compare each extracted statement with its source passage.
3. Confirm negation, resolved issues, proposals and completed actions are distinct.
4. Review a sample of notes with no extracted barrier to find missed information.
5. Restrict access to what each role needs, and record corrections without deleting the history.

Track errors by type: unsupported interpretation, missed statement, wrong record, copied text or incorrect action status. Improve the rules and test them again. A polished summary is not a substitute for these checks.

Frequently asked questions

What should a mentor note include?

A useful starting structure is topics, progress, barriers and next actions, with date, author and record identity. Adapt it to the program and document only relevant information. For actions, distinguish a proposal from an agreement or completed support.

Should emotional statements be removed?

No. A relevant statement from the participant can be recorded accurately and attributed. The problem is inventing or misrepresenting a feeling. Keep reported statements separate from staff interpretation and avoid collecting unnecessary sensitive detail.

Does a quotation make an AI summary trustworthy?

It makes the claim easier to check, but does not guarantee that the quotation supports the summary. Review context, timing, speaker, negation and record identity. Mark missing evidence rather than supplying a plausible answer.

How do we count a recurring barrier?

Define the unit: distinct dated observations, sessions or weeks. Deduplicate copied notes and check whether later information resolves the issue. Separate an author’s statement of duration from a duration calculated from records.

What does a missing note mean?

It means the expected documentation is absent, if the schedule confirms that a note was due. It does not establish that the participant received no support or is about to leave. Check the session and record status.

Can the notes show a program-level problem?

They can reveal a pattern worth investigating. Check whether the statements are current and comparable, whose experience is missing, and what alternatives participants prefer. A recurring theme alone does not prove the cause or the right solution.

Use the notes in reporting

Continue to the job-training funder-report lesson to combine these records with outcomes and coverage. If a monetary valuation is appropriate to your decision, the SROI lesson is an optional analytical step, not a requirement for every program.