Munin AI

Two model services run the product. Deepgram transcribes the call, and OpenAI models fill the adviser's template, extract structured client data, write notes, action items and a summary, draft the follow up email and answer questions in the client profile chat. Professional adds live transcription and live template filling, so an adviser can watch the template complete during the call, and a dynamic checklist binds answers to template items and tracks what has been covered. Past calls can be uploaded and processed the same way. Without the models the templates would sit empty and the client profiles would hold only what an adviser typed, so there is little left to buy.

Every output is a draft that the adviser sends. The follow up email is a draft, notes can be refined after the call, and the workflow ends with the adviser sending the email. Calendar access is read only by design. The privacy policy lists the Google and Microsoft scopes requested and states that Munin never creates, edits or deletes events or adds notes or attachments to them. With no CRM connection there is no automatic write into a system of record. No approval control is described as such, because the product's reach stops at drafts inside Munin.

Transcription, template filling and the profile chat carry no published accuracy figure, test or validation method. Notes are filed under each client with the source meeting, and the adviser can refine them before relying on them. On Professional, the dynamic checklist shows which template items were answered during the call, so gaps are visible before the follow up goes out. The terms place any reliance on the service's content at the user's own risk.

The only performance claim on the site is a meeting turned into a follow up in under five minutes, and no customer, case study or outcome figure is published. In the T3/Inside Information 2026 Software Survey, Munin was one of fourteen tracked AI notetakers, with 0.10 percent market share and a 6.33 user rating from a very small base of users. The site's example questions for the client profile chat, such as a client's children's names or the team they support, show the kind of relationship detail it is meant to recall.

Munin says it configures its AI vendors to use submitted content only to provide the requested service and not to train generalized models. Google data is segregated and never sent to AI or speech to text vendors. Calendar data is deleted or irreversibly deidentified once a feature no longer needs it. How long transcripts and notes are kept, and whether Munin itself trains on meeting content, are not stated. The privacy policy reserves collection of the edits advisers make to AI drafts. Customers can export their data and outputs when they leave.

The privacy policy is specific about connected data. Calendar access is read only, tokens are encrypted, cached calendar data is kept only as long as a feature needs it, and revoking access deletes stored tokens and cache. Google data follows the Google API Limited Use policy and is never used for marketing or sent to AI vendors. Munin does not rent or sell personal information.

It may collect the edits an adviser makes to AI drafted emails, and the policy does not limit what those edits are used for. Users can request access, correction and deletion by email. No reference to GLBA or Regulation S-P and no breach notice commitment is published.

Munin runs on hosting it describes as SOC 2 certified, and its security section lists its own SOC 2 Type II certification as a 2026 goal and ISO 27001 certification as a 2027 goal. The same roadmap adds an annual penetration testing program in 2026 and a public bug bounty in 2027. Controls in place include AES-256-GCM encryption at rest, TLS in transit, hashed passwords, row level security on every database table and isolation between teams. Shared document links carry passwords, expiry dates and download controls, and every document has an activity audit trail.

As a software supplier, Munin holds no license and has no regulator of its own. Its terms draw the advice line: the AI assistant and Munin's staff do not and cannot recommend specific securities or judge what is suitable for an individual, and offer only informational content. No archiving, books and records or supervision feature is described for SEC or FINRA purposes, and no connection to an archiving service such as Smarsh or Global Relay is listed. Advisers keep notes and drafts inside Munin, where each is filed under the client.

Data separation is the one structural safeguard Munin describes. Google data is kept in a separate flow and never reaches the AI or speech vendors, which see only content the user provides directly. No AI governance framework, named owner or fairness testing is published, and nothing addresses how transcription and extraction perform across accents, speaking styles or audio quality.

Under the terms, the user carries the risk. The service is provided as is and at the user's sole risk, reliance on its content is the user's responsibility, and Munin limits its liability for damages under any legal theory. Users indemnify Munin and defend covered claims at their own expense if asked. Disputes go to individual arbitration in Austin, Texas, under Delaware law, with costs and fees to the prevailing party. Nothing promises the accuracy of a note, extraction or draft.

OpenAI and Deepgram are named as Munin's current AI providers, with links to each provider's usage, privacy and security policies. Deepgram's linked policy covers speech to text, and OpenAI supplies the language models. No model version, hosting region for inference or subprocessor list is published, and calling the providers current leaves open that they may change, with no stated notice process. Stripe handles payments.

Connections cover meeting platforms, calendars and files. Munin works with Zoom, Microsoft Teams and Google Meet. It reads Google Calendar and Microsoft 365 calendars through read only scopes that the privacy policy lists by name, including Microsoft's Calendars.Read.Shared, and it never writes events or notes back. Professional ingests Google Drive files and bulk imports of past calls.

No CRM, planning tool, custodian or archiving connector is listed, so notes and data stay in Munin unless an adviser copies them out. Data and outputs can be exported when a customer leaves, with help from Munin.

Services are hosted in the United States on infrastructure Munin describes as SOC 2 certified, and the privacy policy tells users in the UK, the EU and elsewhere that their data is processed there. The hosting provider and region are not named. Data is isolated between teams with row level security on every table. Munin is used in the browser alongside Zoom, Teams and Google Meet, and past calls can be uploaded for processing. No private or on premises deployment is offered.

Essentials costs $39 a month, or $29 a month billed annually, and covers templates, client profiles with AI chat, uploads, calendar integrations and post meeting outputs. Professional costs $79 a month, or $59 annually, and adds past call uploads, live transcription, a dynamic checklist, Google Drive imports and priority support. Enterprise is quoted by email for larger teams. Checkout runs through Stripe. Monthly plans can be cancelled with access to the end of the month, annual plans run to the end of the year, and upgrades are available at any time.

Financial advisers are the stated market throughout the site, with per user plans for individuals and an enterprise option for larger teams with special requirements. No firm type, size range or named customer is given beyond that. Access is screened. Registration asks for a name, email address and LinkedIn profile, the terms say every applicant's identity is verified, and a phone interview may be required before an account is granted. Services are hosted in the US, and users elsewhere are told their data will be processed there.