Microsoft Teams thread
JLJordan Lee
“Sam knows about the internal project, he can walk you through it.”
Mira: Supervision Agent
The AI supervision agent built on your firm's own handbook
Mira understands the context of conversations, explains why a message needs attention, and routes it to the right reviewer. Your team can focus on real risk signal while keeping every review decision in human hands.
Written Supervisory Procedures
Kelderwick Partners
3.1Gifts and entertainment
4.2Promissory statements1
Communications may not state or imply that a result is certain.
4.5Off-channel communication
Item 88211
Dana Reed to Ellen Cole · Email
ReadingFlaggedDana ReedOct 14 · 10:15 AM
Honestly this one can’t lose. I’d put the whole rollover in it.1
AIMira’s reasoning
Item 88211
- Read in context, across the whole thread
- Matched to handbook clause 4.2, Promissory statements
The message tells a client the investment cannot lose, so it promises an outcome.
Severity: HighPromissory statements · 4.2
Review queue
Branch 114
- Routed to Marcus OkaforBranch 114 supervisorMOAwaiting reviewer
What firms see with Mira
98%
fewer false positives with Mira
80%
reduction in communications review time
The Problem
Keyword-based review can leave teams sorting through noise.
Lexicon alerts can fill review queues with messages that don’t need attention. Teams also have to maintain the lexicon as policies change, and a flag alone may not explain why a message needs review.
How Mira works
From your handbook to a program your reviewers control
See how Mira applies your policies to captured messages and where your team stays in control.
01
Your handbook becomes a policy program you approve
Upload your communications policy handbook, and Mira proposes a versioned policy program for your team to review. Its proposals draw on SEC and FINRA rules, guidance, and enforcement history since 2010.
- Each policy carries the handbook section it came from.
- Nothing is enforced until your team approves the program.
- The program is versioned, so every change is identifiable.
- Each firm decides who can approve a change.
One policy from Policy program v1 at Kelderwick Partners, Promissory statements, traced to its sources: the handbook, Written Supervisory Procedures, section 4.2; the rule, FINRA 2210; and guidance and enforcement history, considered. Version 1, approved by Paula Nash. Changes need approval from Compliance.
Policy program v1Kelderwick Partners
Handbook
3.1Gifts and entertainment
4.2Promissory statements
Communications may not state or imply that a result is certain.
4.5Off-channel communication
Written Supervisory Procedures, section 4.2
Promissory statements4.2
Draft
Awaiting approval
Version 1 · Approved by Paula Nash
- RuleFINRA 2210
- Guidance and enforcement historyconsidered
Changes need approval from: Compliance
02
Focus review on messages that may need attention
Mira assesses captured messages against your firm’s policies, considering the context of the conversation rather than just matching keywords.
- The whole thread read, with the people it names.
- Clean messages stay out of the review queue.
- Each flag classified by policy area and scored for severity.
- Each flag routed to the right reviewer.
Two example messages from Branch 114 on Microsoft Teams, each read in context. Jordan Lee mentions that Sam knows about an internal project: no flag, and the message stays out of the queue. Dana Reed names a company ahead of a dated announcement: flagged under Insider information 6.3 with severity high, and routed to Marcus Okafor, Branch 114 supervisor.
Branch 114
Review queue
No flag
stays out of the queue
Microsoft Teams thread
DRDana Reed
“Quietly, load up on Kestrel Therapeutics before the Nov 12 announcement.”
Flagged · Insider information 6.3 · Severity: High
MORouted to Marcus Okafor, Branch 114 supervisor
03
Improve policy logic with reviewer feedback
Mira groups similar review decisions and proposes refinements to the policy program.
- Refinements arrive as drafts, never as live changes.
- A reviewer approves each versioned update before it ships.
- Each firm sets who can approve a change.
- The program changes only through that approval.
A proposed update to Gifts and entertainment 3.1, version 1 to version 1.1. The old logic, flag any mention of event tickets, is struck through. The new logic, flag event tickets above the firm's gift limit, is highlighted. It was drafted from reviewer decisions on similar flags: dismissed by Marcus Okafor, conference tickets within limit, and escalated by Marcus Okafor, tickets above the gift limit. The update stays a draft, awaiting approval from Paula Nash. If Paula Nash approves, it goes live as version 1.1; if she rejects it, it does not.
Gifts and entertainment 3.1
Drafted from reviewer decisions on similar flags
- Dismissed by Marcus Okafor · conference tickets within limit
- Escalated by Marcus Okafor · tickets above the gift limit
Proposed update · Gifts and entertainment 3.1v1 → v1.1
Flag any mention of event tickets
Flag event tickets above the firm’s gift limit
Draft
v1.1 live
PNAwaiting approval · Paula Nash
PNApproved · Paula Nash
PNRejected · Paula Nash
04
Case management
Investigations are a query, not a project.
Bring a focused investigation into one case. Mira helps your team find relevant evidence in the captured messages you choose, while your team reviews the findings.
- Start with a requestdefine the scope of your case.
- Ask, don't searchplain-English questions run across the people, channels, and dates in the case.
- You decideyour team flags, clears, or records every finding, with a full activity history.
- Hand it overexport the full case, or a summary PDF report Mira drafts for your review.
The case Cole account review, opened October 20, 2026 by Paula Nash. The uploaded file Legal request.pdf resolves into the case setup: a date range, the person Dana Reed, the channels text and email, and a playbook. Inside that case, the question Find off-channel communications returns item 88231: Dana Reed to Ellen Cole, by email, who writes Text my personal cell instead, it’s easier, with the risk category Off-channel communication. Outside the case, Paula Nash marks it Flagged, the activity history logs the action, and a Summary report is ready as a draft for review.
Cole account reviewOpened Oct 20, 2026 · Paula Nash
Legal request.pdf
- Date range
- People: Dana Reed
- Channels: Text, Email
- Playbook
Find off-channel communications
Item 88231Dana Reed → Ellen Cole · Email
“Text my personal cell instead, it’s easier.”
Off-channel communicationPNFlagged
Paula Nash marked item 88231 Flagged
Summary report · Draft for review
05
Reports and dashboards
Describe the dashboard. Get the numbers.
Convert a query into a report. Mira takes supervision activity and turns it into a report that helps your team see patterns and answer questions.
- Ask in plain Englishadd, rearrange, and filter views by team, person, or date.
- Explore the detaildrill into messages behind your results.
- Run, save, or scheduledaily, weekly, monthly or quarterly, as PDF, or CSV.
- Review historyexplore supervision activity over time.
The prompt Show review outcomes for Branch 114, last 90 days builds a dashboard from supervision metrics: bars of flagged messages by month and a status breakdown of dismissed, escalated, and confirmed. Selecting the October bar opens the messages behind it: item 88214, Gifts and entertainment, and item 88231, Off-channel communication. The report is scheduled monthly as a PDF.
Review outcomesBranch 114
Show review outcomes for Branch 114, last 90 days
Flagged, by month
- Aug
- Sep
By outcome
- Dismissed
- Escalated
- Confirmed
Messages behind this bar
- Item 88214Gifts and entertainment
- Item 88231Off-channel communication
Show the reasoning behind a supervision decision
When an examiner asks why a message was flagged and what happened next, the record connects Mira’s reasoning with your team’s review. Red Oak calls this Compliance-Grade AI™.
- 1
Explainable findings
Every flag carries Mira's reasoning and the evidence behind it.
- 2
Policy-based controls
Each flag traces to the policy and handbook clause that triggered it.
- 3
Human oversight
A person makes every review decision, and the program changes only through approved updates.
- 4
An auditable decision history
Every review action lands on one audit trail with the name of the person who took it.
The evidence trail for item 88211, escalated, its stops numbered to match the four parts of Compliance-Grade AI™: the policy, Promissory statements, and its handbook clause, section 4.2; the context, a WhatsApp thread with a client about an IRA rollover, with a prior coaching note under the same policy; Mira’s reasoning; the decision, escalated by Marcus Okafor; and the audit trail, from 10:31 when Mira flagged the item under section 4.2 to 10:42 when Marcus Okafor escalated it to Paula Nash.
Case summary
Item 88211Escalated-
Handbook clauseSection 4.2
-
Rep historyA prior coaching note under the same policy
-
“The message tells the client the investment cannot lose and recommends moving the whole rollover into it, which reads as a performance promise under section 4.2.”
-
- 10:31 Mira flagged item 88211 under section 4.2.
- 10:39 Marcus Okafor opened the full conversation.
- 10:42 Marcus Okafor escalated item 88211 to Paula Nash.
Trust center for access, security, and data handling.
Recognition and customer feedback
8 consecutive years
on the Inc. 5000
84
Net Promoter Score
Visionary
2026 Gartner Magic Quadrant for Digital Communications Governance and Archiving
Explore supervision and Red Oak’s AI approach
Communications supervision
Supervision built on your own compliance handbook, with every flag routed.
Read moreAI approach
How Red Oak governs AI, and what Compliance-Grade AI™ means.
Read moreWorkflows
Review queues with named owners, reviewer groups, and separation of duties.
Supervision
The product hub for communications and website supervision.
Read more
Mira connects to the live channels you capture. Start with one review group or one channel, then widen when your team is ready. Access runs on single sign-on with the permissions your administrators manage, and records kept on Red Oak sit in WORM storage. Trust center
See how Mira supports your supervision team
Explore how Mira turns policies into a review program, assesses a message, and explains a flag for your team to review.
Mira, answered
-
An AI supervision agent reviews a firm's captured communications against its supervisory policies, flags what needs attention, and explains why, while a person makes every review decision. Mira is Red Oak's supervision agent. It reads your policies and procedures, builds risk categories from them, and checks each captured message against them. Communications supervision
-
Mira builds its supervision policies from your firm's own compliance handbook. Upload the handbook, and Mira turns each supervision rule into a versioned policy program that your team reviews and approves. The policies it proposes draw on SEC and FINRA rules, guidance, and enforcement history since 2010. See a demo
-
Mira learns from your team's review decisions, and every update it drafts waits for a reviewer's approval. Each dismissal, escalation, or confirmation informs Mira, which groups similar decisions across the firm and drafts refined policy logic. The refinement ships as a versioned update once a reviewer approves it. Read our approach to AI.
-
Each flag's case summary shows the policy that triggered it, the handbook clause it derives from, the context, the rep's history, and Mira's reasoning, then the reviewer's decision. Every classification, routing step, escalation, and resolution is logged automatically, so each decision has a record that is easy to follow. Read our approach to AI.
-
See the trust center.