Skip to content
BlogCompliance

How Mesh flags compliance language, and why it starts off

Eight categories, three severities, the quoted passage, and your decision on each flag. And why compliance scanning stays off until you turn it on.

4 min read

Off by default: Scanning starts when you turn it on

Compliance scanning in Mesh is off until you turn it on. The switch is in Settings, under Compliance, and it reads “Scan meetings for compliance flags.” While it is off, Mesh writes no compliance summary and no flags. The meeting page shows no Compliance card, and the meeting’s review page says “Compliance review is off.” Your summary and action items arrive as usual.

Starting off means nothing new lands in your review until you ask for it. A flag is AI output, and every flag asks you for a decision. That is worth it when you want a second look at the language of each meeting. It is your call, not a default you have to find and undo.

Turn it on, and Mesh also writes a short compliance summary of each meeting it processes. A meeting with no flags shows “No compliance flags detected” in the Compliance Flags panel. Flags come from a review that reads the redacted copy of the transcript, with client names replaced by placeholders. Here is what redaction removes, and where it stops.

Fig. 1 Scanning off, and scanning on
What Mesh does for each meeting with compliance scanning off and on
For each meetingOffOn
Summary and action itemsWrittenWritten
Compliance summaryNot writtenWritten
Compliance flagsNoneWhen the review finds something
The meeting’s review page“Compliance review is off”The Compliance Flags panel
Settings, Compliance: “Scan meetings for compliance flags.”

Categories: Eight kinds of language, three severities

Every flag carries one category and one severity. The categories are the labels you see on each flag. The definitions in Fig. 2 come from the instructions Mesh gives the AI on every review, shortened to a line or two each.

Fig. 2 The eight categories, and the severity rules in the AI’s instructions
The eight compliance categories, what each covers, and the severity rule that Mesh’s instructions give the AI
CategoryWhat it coversSeverity rule given to the AI
SuitabilityA recommendation that may not fit the client’s stated risk tolerance, objectives, time horizon or financial situation.Set per finding
Disclosure GivenA required disclosure that was given: a risk explained, a fee disclosed, a conflict acknowledged.Always Info
Disclosure MissingA required disclosure that was not given when it should have been: fees, risks, conflicts of interest, material limitations.Set per finding
Risk DiscussionHow investment risks were explained, or no risk discussion next to a recommendation.Set per finding
Fee DiscussionHow fees, costs and expenses were discussed, or vague or missing fee talk next to a recommendation.Set per finding
Guarantee LanguageLanguage that implies guaranteed returns, assured outcomes, certain performance or promised results.Always Action Required
Complaint DetectedThe client expresses dissatisfaction, frustration or confusion about advice, raises a concern, or complains.At least Warning
Recommendation MadeA specific product, strategy, fund or allocation the advisor recommends. The severity depends on whether suitability was established first.Set per finding
Labels as they appear on a flag. Definitions and severity rules condensed from the instructions Mesh gives the AI. Mesh does not check the AI’s severity against these rules.

Three categories carry a fixed rule in those instructions. A disclosure that was given is always Info, because it records good practice. Guarantee language is always Action Required. A complaint is at least a Warning. For the other five, the AI sets the severity from the context of each finding. These rules are instructions to the AI. Mesh does not check the severity the AI picks against them.

The three severities mean this:

  • Info: routine and handled. No action needed.
  • Warning: it needs your attention, or more documentation, before you finalize.
  • Action Required: a possible regulatory concern for a compliance officer to review.

Your decision: Every flag waits for you

Flags appear in the Compliance Flags panel on the meeting’s review page. Each one shows the severity, the category, a short description, the passage the AI quotes from the transcript, and who said it when the AI can tell. The quote is the AI’s, so check it against the transcript. Select a flag, and Mesh opens the transcript and marks the passage when it can find it. Filter the list to All, Warning+ or Action Required. The same flags also show in the Compliance card on the meeting’s Review tab, where you can read them but not review them.

Fig. 3 A compliance review, as it appears in Mesh

Compliance Flags

121
InfoDisclosure Given

Advisor explained that a Roth conversion is taxable in the year it is made.

“Just so you know, the converted amount counts as income for 2026.”
Dana Whitaker
WarningRecommendation Made

Conversion amount recommended. Document the tax-bracket rationale before you finalize.

“I'd keep the conversion under the top of your current bracket this year.”
Dana Whitaker
WarningRisk Discussion

Client raised concern about market volatility before retirement.

“I can't stomach another year like 2022 right before we retire.”
Robert Caldwell
Action RequiredGuarantee Language

Language may imply a guaranteed outcome. Review before it reaches the client file.

“This approach basically guarantees you won't run short.”
Dana Whitaker

AI-generated compliance flags are advisory only. They do not constitute legal advice or regulatory determinations. All findings require human review.

Four flags from a fictional annual review, in the real panel. Open Review on a flag to try the actions. Nothing here is saved.

Open Review on a flag, add a note if you want one, and choose one of three actions.

Fig. 4 The review actions
Acknowledge
Use it when the flag is fair and you have handled it. The card then reads “Acknowledged.”
Dismiss
Use it when the flag is not a concern. The card reads “Dismissed.”
Escalate
Use it when the flag needs a closer look, for example from a compliance officer. The card reads “Escalated.” Mesh marks the flag. It does not send it to anyone.
Notes
Optional. Mesh stores your note with the flag, encrypted.
The record
Mesh stamps who decided and when on the flag, and writes the review to the audit log.

After you decide, the Review button goes away and the card shows your decision. Mesh stores the description and the quote of each flag encrypted.

The limits: A flag supports your review

Under its list of flags, the Compliance Flags panel on the review page says this:

AI-generated compliance flags are advisory only. They do not constitute legal advice or regulatory determinations. All findings require human review.

Two things follow. A flag is a reason to look, not a determination. And a meeting with no flags is not a clean assessment. The review instructions tell the AI to flag only genuine concerns and never to invent findings to look thorough. So a quiet meeting can still hold something that you would catch.

Mesh does not replace your archive either. A recap email goes to a client only after you approve it, and it goes out from your own mailbox, so your firm’s existing archiving and supervision capture it. What Mesh adds is the record around the meeting: each flag, the passage it quotes, and your decision on it.