Skip to content

BRK482 Breakout Tue 17 Nov 15:15 PST

FinOps for AI: Partner Playbook for Building AI Solutions and Services

Matt Kerner, Adam Bogobowicz, Jimmy Priestas

Session description

Microsoft's catalogue description

AI is creating unprecedented opportunities for partners, but success requires more than innovative solutions. It requires managing the economics behind AI. In this session, Microsoft leaders and industry experts share a practical FinOps framework for AI to help partners anticipate and govern costs, optimize token consumption, and build repeatable, profitable services. Hear real-world partner examples and leave with actionable guidance to turn AI cost management into a competitive advantage.​

Demo timestamps

Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.

  1. Skip to 0:09:55

    Developer: Microsoft Foundry cost management

    Foundry cost page shows $124 over 7 days across 5 agents. Drills into traces and cost drivers, compares models in an evaluation run, and runs an optimiser (about $0.02 vs $0.19 baseline per run). Sets budget alerts and blocks.

  2. Skip to 0:19:56

    IT pro: Agent 365 spending policies

    In the M365 admin center, shows credit requests and top users. A spending policy for a Research group sets $100,000 monthly and $1,000 per user, with model choice, billing options and delegation roles.

  3. Skip to 0:26:52

    Business leader: value and forecasting

    Spend forecast with a cone of uncertainty against budget. AI prompts are classified into about 7,000 kinds of work to estimate value and ROI by use case. Natural-language usage questions are supported.

  4. Skip to 0:30:38

    End user: Cowork credit gauge

    A credit gauge shows reset date, per-day activity cost, a '20% of monthly limit remaining' alert and a request-more-credits flow. Agent 365 and Foundry policy connection is said to arrive early in 2027.

Watch on-demand (opens in a new tab)

Resources named in the session

← All sessions