Deep AI Insights from Collecting Teams Meeting Media with Numonix CEO Michael Levy

Prefer audio?

Who should listen: Microsoft 365 service owners, compliance and risk teams in regulated industries, and UC architects evaluating whether to buy or build Teams media capture for AI, analytics or compliance use cases.

Guest: Michael Levy — CEO, Numonix

Michael Levy, CEO of Numonix, a Microsoft-certified compliance recording vendor, explains how the same media-capture bot infrastructure is being repurposed as a headless 'capture as a service' layer for AI platforms and in-house analytics builds. The conversation covers where this sits relative to Copilot, why building your own compliance recording bot is harder than it looks, and the commercial upside of transacting through Azure Marketplace.

Many thanks to Numonix, sponsor of this episode.

Key insights

• Numonix has split its compliance product into a UI-less 'media acquisition' offering: customers get an API and a storage account in their own tenant with the media file plus XML metadata, rather than paying for front-end features they will never use. ▶ 3:32
• Recording scope can be driven by policy rules rather than blanket capture — internal-only, external-only, or triggered by keywords in the calendar invite (record only if the subject contains 'record', or suppress if it contains 'private' or a do-not-record code) — which limits the data surface before any AI analysis happens. ▶ 4:04
• Gong uses Numonix as its Teams capture layer for revenue intelligence, an example of AI vendors buying media acquisition rather than building compliance bots themselves. ▶ 5:05
• Deployment is policy-based and hardware-free: admin consent URL, bot registered as an enterprise app, a compliance recording policy assigned to a security group, then the bot joins as an invisible participant and drops the file where you choose. ▶ 7:39
• The strongest argument against rolling your own bot is Premier support access to Microsoft's recording group — without it, diagnosing whether a fault is Graph, Teams or an AD group policy issue is very difficult, and API changes such as bot grouping and delta rosters for large meetings require continuous testing. ▶ 8:42
• Levy claims the consumption-based capture service costs less than hiring a single developer to build and maintain the integration, and that customers are starting to record now purely to build a legacy data set for AI projects they will not begin for six months. ▶ 6:07
• Copilot is a per-user personal assistant licence that cannot be shared across a team; a capture-plus-Cognitive-Services pipeline lets you generate shared meeting summaries — for example, an automatic all-hands summary posted to SharePoint with expiry-based sharing controls. ▶ 13:18
• Selling through Azure Marketplace draws down pre-committed Microsoft spend, appears as a line item on the Microsoft bill and bypasses vendor onboarding that can take three months in large enterprises; being Azure-native also lets you answer data-centre security sections of an RFP with 'Microsoft Azure'. ▶ 16:22

Insights summarised by AI from the episode transcript, reviewed by the Empowering.Cloud team.

Listen: Apple Podcasts · Spotify · Other platforms