How Fellow's Advanced Trackers Flag MNPI, Insider Information, and Promissory Language in Meetings

Most firms now record hundreds of meetings a week, yet very few have a realistic way to supervise them. A compliance team cannot listen to every client call, deal review, and internal huddle. So risky language (an offhand hint about unannounced earnings, a promise of guaranteed returns, a personal email address used to move files) sits in a transcript nobody reads. AI meeting trackers are built to close that gap.

Instead of sampling a handful of recordings or relying on what individuals choose to share, you describe what matters and let AI review every recorded meeting against it. This guide covers what AI meeting trackers are, why keyword lists fall short for nuanced risk, how Fellow's Advanced Trackers work step by step, and how to set them up so reviewers spend their time on real issues instead of noise.

What are AI meeting trackers?

AI meeting trackers are rules that automatically scan meeting transcripts for specific topics or types of language, then flag each match for review. They turn a library of recordings into a searchable signal layer, so the conversations that need attention surface on their own.

Advanced Tracker: a Fellow tracker that pairs a plain-language prompt describing what to flag with optional labels and meeting rules, and always runs on Fellow's most accurate AI model.

Advanced Trackers are part of Fellow, the secure AI meeting assistant that turns every meeting into shared, searchable intelligence. Fellow captures conversations across Zoom, Google Meet, Microsoft Teams, in-person meetings, and Slack huddles, with a bot or without a bot joining the call. Every recorded meeting with a transcript can then be checked against your active trackers, which is why trackers sit at the center of how regulated firms supervise meetings in Fellow.

Why do keyword lists miss real compliance risk?

Keyword lists catch exact words, not meaning. Risky conversations rarely use the vocabulary a compliance team predicts, so keyword monitoring produces both missed matches and a pile of false positives.

Consider a banker who tells a client, "the numbers we report next month will look a lot better than the street expects." That sentence never says "earnings" or "guidance," but it is exactly what an MNPI policy is written to catch. Meanwhile, a keyword list for "acquisition" fires on every meeting where someone discusses acquiring a new software license.

Fellow offers three tracker types, each suited to a different level of nuance:

The difference matters most when the cost of a miss is high. A Concept tracker learns from a few example phrases, while an Advanced Tracker reads your full description of the risk, so it can catch a paraphrase that no example list anticipated.

How do Fellow's Advanced Trackers work?

Compliance officers can describe what they want flagged in plain language, and Fellow's AI reviews every recorded meeting against that description. Each match is saved with its place in the transcript, a confidence score, and a label, ready for review.

Setup takes a few minutes on the Trackers page, where you pick Advanced and follow a short step-by-step flow:

1. Name the tracker and write the prompt. Describe the risk in your own words. For example: "Flag any discussion of material nonpublic information, including financial figures before earnings, unannounced M&A, or undisclosed product launches."

2. Add labels (optional). Labels are sub-categories, each with a one-line description, such as "Pricing: discounts, list price changes, or contract terms." Every match is sorted into exactly one label, which turns a flat list of matches into categories your team can triage, route, and report on.

3. Set rules (optional). Choose which meetings the tracker runs on, by title, attendees (a specific person, internal or external participants, an email, or a team), organizer, meeting length, or attendee count.

4. Refine with AI help. One click turns a rough idea into a structured prompt, a suggested name, and 2 to 6 clean labels. You review the changes before applying them, so no prompt-writing expertise is required.

5. Let analysis run automatically. Every recorded meeting is checked against your active trackers, with no one assigned to listen or sample.

6. Review the matches. Matches appear on the meeting recording itself and can be used to filter your meeting library. For compliance teams, they also flow into Fellow's compliance portal for structured review.

Because every match points to an exact moment in the transcript, reviewers jump straight to the relevant exchange instead of scrubbing through a full recording.

What can you track with Advanced Trackers?

Advanced Trackers work for any topic you can describe in a sentence or two. In financial services, these trackers map directly onto daily supervision work. Hedge funds and investment banks can run an MNPI tracker across external calls, while private equity firms can point a tracker at meetings with portfolio company management.

Advanced tracker examples for finance and regulated financial services

Can AI meeting trackers work outside finance?

Yes. Advanced Trackers work for any risk you can describe in a sentence or two, so the same setup that flags MNPI on a banking call can flag patient information or risky clinical language in a healthcare or healthtech meeting. The mechanics stay the same: write the prompt, add labels, scope the tracker with rules, and review each match as a signal.

Healthtech and healthcare teams face their own version of the MNPI problem. A sales call with a hospital customer can drift into patient details, a support call can end with someone reading out a member ID, and a product demo can imply that a tool diagnoses or treats a condition. None of these follow a predictable word list, which is why describing the behavior works better than listing keywords.

Fellow is SOC 2 Type II and HIPAA compliant, so these trackers run inside a platform built for regulated teams. As with finance, a match is a signal for human review, and your compliance team decides whether it is a real issue.

Advanced tracker examples for legal, healthtech, sales, customer success, and HR

How do compliance teams review tracker matches?

Tracker matches flow into the compliance portal, one place to search, review, flag, and audit every meeting recorded in Fellow across your workspace. Reviewers see meetings independent of individual sharing settings, so oversight never depends on what someone chose to share.

A typical daily review looks like this:

1. Open a saved view. Filter by tracker, participant, or date range, then save recurring reviews (for example, "MNPI, external meetings") as named views for daily or weekly monitoring.

2. Filter by confidence. Start with the highest-confidence matches so the most likely issues get attention first.

3. Jump to the moment. Each match links to its exact place in the transcript and recording.

4. Collaborate on the call. Move matches from flagged through to closed, and leave comments for your team.

5. Dismiss false positives with a reason. Dismissal reasons are logged, so the decision to close a match is documented, not silent.

Every review action (reviews, comments, dismissals, edits, and redactions) is logged, and redactions are always visible rather than silently removed. The result is a review process your firm can stand behind in an exam or audit. Access to the portal is granted only by a compliance admin, and sensitive trackers can be locked to the portal so only authorized reviewers ever see them.

How do you write an Advanced Tracker that catches the right things?

The best Advanced Trackers are narrow and specific: one risk per tracker, concrete examples in the prompt, labels that mirror how your team triages, and rules that limit the tracker to meetings where the risk can actually occur.

Use these practices when you build your first trackers:

1. Describe the behavior, not the vocabulary. "Flag any promise or implication that a client will earn a specific return" catches far more than a list of words like "guarantee."

2. Include examples inside the prompt. Naming the forms a risk takes (earnings figures, unannounced deals, undisclosed launches) sharpens what the AI looks for.

3. Design labels around your workflow. If MNPI tied to M&A goes to a different reviewer than MNPI tied to earnings, make them separate labels.

4. Scope with rules. A tracker for promissory language probably belongs on external client meetings, not internal engineering standups.

5. Start rough, then refine with AI help. Type your rough idea, let AI help structure the prompt and labels, and review every suggested change before applying it.

Common mistakes to avoid

• Writing a keyword list as a prompt. A prompt that just reads "earnings, M&A, merger, deal" gives up the nuance that makes Advanced Trackers worth using.

• Overlapping labels. Each match lands in exactly one label, so two labels that describe the same thing split your results unpredictably.

• Running every tracker on every meeting. Broad coverage raises false positives and uses up your active-tracker limit. Rules keep each tracker focused.

• Treating a match as a finding. A match is a signal for human review, with a confidence score attached. Your reviewers make the call.

What security and governance controls work alongside trackers?

Trackers are only as trustworthy as the platform around them. Fellow pairs Advanced Trackers with firm-level controls your compliance team configures once and applies across every team and business unit.

• Sensitive trackers: only compliance admins can create or edit trackers marked sensitive, and regular users cannot see them.

• Configurable data retention: set auto-deletion schedules at the workspace level, keep AI summaries longer than raw transcripts, or choose a zero-day retention option so source recordings and transcripts are not retained after processing.

• Pause, resume, and redaction: attendees can pause recording during a sensitive discussion, and teams can redact sensitive content after the meeting. These sit alongside Fellow's broader privacy controls.

• Recording restrictions by rule: on Enterprise, admins can block or require recording based on meeting title keywords, attendee emails, or internal versus external attendees.

• Consent and sharing controls: capture consent for auto-recorded meetings and password-protect shared recap links.

• Access permissioning: a permission-controlled meeting library, plus SSO, SCIM, and role-based access control for complex org structures.

• Certifications: SOC 2 Type II, HIPAA, and GDPR compliant, with AES-256 encryption.

• Zero training on your data: Fellow never trains AI models on your notes, transcripts, audio, or summaries, on any plan.

• Archiving and extensibility: the Global Relay integration supports records management, alongside 50+ native integrations, the MCP Server, and the Developer API.

Fellow supplies the controls, and your firm decides how to configure them for its own compliance program. Fellow's compliance features are designed to support that program and do not constitute legal or regulatory advice. Some capabilities, such as transcript redaction, depend on your plan and configuration.

Conclusion

Recording every meeting is easy. Supervising every meeting is the hard part, and keyword lists were never built for the way people actually talk about earnings, deals, or guarantees. Advanced Trackers let you describe the risk in your own words, review every flagged moment with a full audit trail, and back it all with retention, access, and redaction controls your firm configures. To see how the pieces fit together, explore Fellow's compliance features alongside the rest of its AI meeting notes platform.