Each day has a theme. Tuesday is the platform: governance, semantic views, Iceberg, compute, apps and observability. Wednesday is marketing and sales, and Thursday is finance. The platform day is the one for this site.

• State of Observability: Telemetry as a Data Asset (Tue 17:30 EST, Jacob Leverich). The pitch is to move telemetry out of special-purpose stores and into the general data estate. That makes telemetry one more table that is late by however long ingest takes. Ask what age the investigation UI shows next to each value. Every View Is Late makes that the first of five rules for any live view. See ObservabilityCON 2026 for Grafana's approach, where Adaptive Telemetry cuts cost by controlling what gets ingested, and Agent Telemetry Systems.
• Secure and Govern Data & Agents Across Every Surface (Tue 14:00 EST, broadcast). Covers agent identity, access control, behavior auditing and cost. Read it against A Practical Layout for Agent Identities and the gateway version of the same question in the Postman agent gateway webinar. Bot and Agent Trust Management covers the four agent registries shipped in 2026, none of which interoperate, and Agent Permission Guardrails separates session-wide grants from per-operation checks.
• Build One Business Context for Every Agent, App, and Person (Tue 14:30 EST). Cortex Sense, Semantic Views and a "business ontology" that declares which metric definition is official. Compare it with the semantic-layer and ontology arguments at Semantic Layer Summit 2026, where Snowflake's Josh Klahr sat on the open-semantics panel, and Microsoft's version of the same move, Fabric IQ binding an ontology to warehouse tables, in Frontier Transformation Week.
• Building the Sovereign AI Stack (Tue 15:30 EST, Satya Nadella and Sridhar Ramaswamy). A fireside chat on what an enterprise must own when it uses frontier models. Watch for whether "own" means weights, traces or evaluation data.
• How to Manage Costs with Tokenomics and FinOps (Thu 15:00 EST) and Attribute and Optimize AI & Cloud Costs in Snowflake (Thu 17:00 EST). The second describes Snowflake's own attribution pipeline: ingest from each provider, a dbt attribution model and anomaly detection before close. Compare with LLM Cost Monitoring.

Each row pairs a session with a note on this site that covers the same problem. A session can appear more than once. Sessions with no match are left out.

Eighteen of the 44 sessions have a match: three strong, eight partial and seven tangential. The other 26 have none. Thursday is the weakest fit, with three of fourteen sessions matched and none strong; ERP, tax, close and the finance case studies have no counterpart here. Wednesday matches nine sessions, but mostly at the edges.

Times are Eastern (EST). Speakers are listed where the agenda names them.

• The business-context session says agents answer "confidently, but inconsistently" when five semantic views define one metric. Once an ontology declares the official definition, does the agent's answer name which definition it used, and how old the data under it is?
• Horizon Catalog promises that "every agent inherits security the moment it's deployed." Is the grant checked on every query, or fixed when the session starts? What revokes it mid-session?
• If telemetry lives in the warehouse, what is the ingest lag at P99, and does the observability UI show it?
• The cost-attribution pipeline catches anomalies "before close." What is the detection delay in days, and how does it compare with the provider bills it reconciles against?

• Agent Telemetry Systems — token-spend ledgers and routing economics for one operator; the small-scale version of Thursday's cost-attribution talk.
• LLM Cost Monitoring and Local-First Harness — token economics and the API-versus-local breakeven.
• A Practical Layout for Agent Identities — role-scoped keys per agent, the identity half of the governance session.
• Identity Store Enterprise Ontology — a declared model checked with SHACL; the model is identities, not metrics, so it shows the mechanism only.
• Bot and Agent Trust Management Software — agent registries, agent SSO and non-human identity in 2026; Horizon's agent governance belongs in the same survey.
• Agent Permission Guardrails — session-wide grants versus per-operation checks, the question behind "every agent inherits security."
• Every View Is Late — five rules for live views, starting with showing the age of the data.
• Declared vs. Effective State — the gap between a declared policy and what the running system does, which is the gap agent governance has to close.
• Semantic Layer Summit 2026 and ObservabilityCON 2026 — the same two arguments from other vendors.

Notes pending.