Pops and Steele tackle a problem most IT and finance leaders haven't caught up to yet: AI subscriptions and API spend are becoming the new shadow IT. Teams are subscribing to AI tools on corporate cards with zero governance or intake process, token-based consumption is replacing predictable licensing, and nobody's tracking the "blast radius" of what these tools can access. They walk through why AI spend breaks the old software asset management playbook (or doesn't — they debate this), how to treat AI subscriptions like governed IT assets, what review cadence makes sense (monthly until you understand it, then quarterly), and how to open the budget conversation with finance before finance opens it with you. Along the way: a $2,000-in-24-hours cautionary tale, a Lexis Nexis printer-page analogy, and Steele's line that becomes the episode's thesis — "whoever ignores the budget conversation loses the seat at the table."

Key Takeaways

• Shadow AI hides wherever there's no intake and governance process. If you don't have a formal way to bring AI tools in, someone already brought one in without you.
• AI spend is jagged and consumption-based, not predictable like traditional per-seat licensing — auto-approval settings and "set it and walk away" agent usage can generate runaway bills.
• Wrong tool, wrong job costs money. The $2K/24-hour example: using a frontier model for basic tasks is "driving a Ferrari to the grocery store."
• Treat AI subscriptions like governed IT assets — track what models, datasets, skills, and prompts are in use, ideally surfaced first in a spreadsheet, then a CMDB.
• Ownership is shared, but accountability isn't optional — the business owns the risk, but the team that brought a tool in without process owns the consequences.
• Review cadence: monthly until you fully understand a new tool's usage pattern; only then step back to quarterly/biannual. New capabilities may need daily/weekly checks at rollout.
• Bring numbers to finance before they ask. Go in with an annual spend estimate, a value narrative, and a straight answer — don't let finance find out from the invoice.
• Read the fine print on new model releases — usage multipliers (like a new model consuming 50% more of your quota) can quietly blow through limits.

Keywords / Tags
AI spend management, shadow AI, shadow IT, AI governance, ServiceNow ITAM, IT asset management, AI subscription tracking, token consumption, CMDB, AI budget, finance and IT alignment, agentic AI risk, AI cost governance, enterprise AI adoption, frontier models, AI ROI, IT leadership, CAB governance, AI sprawl, security risk AI tools, AI procurement

Is AI spend already a line item at your shop — or are you still finding out from the invoice? Drop a comment.

If you're the one who has to explain the AI bill to finance, hit subscribe — this is the show for you.

Tag the person on your team who needs to see this before the next invoice lands.

Want the one-sheeter framework Pops uses to pitch new AI tools to finance? Link in bio.

Support the show