On this episode of Masters of Data, Adam White and David Girvin dig into Sumo Logic’s freshly launched compliance apps for Claude, ChatGPT, and LiteLLM, and why your IT team will want to pay attention before the token bill arrives. We unpack how enterprises can move beyond the “AI black hole” era of shadow IT and actually get eyes on who is using what, how much it is costing, and whether any of it is moving the needle. From Uber burning through its annual AI budget in a few months to Microsoft watching AI costs outpace headcount, we use some instructive real-world cautionary tales to make the case for sanctioned, observable AI deployment. This episode is a must-listen for IT leaders, security teams, AI program owners, and anyone whose job now includes the phrase “show me the ROI.”
0:00 — Introductions
0:33 — Topic overview: Claude, ChatGPT, LiteLLM & compliance apps
1:03 — LiteLLM, OpenAI & Claude compliance apps explained
2:19 — What the compliance apps actually cover (spend, identity, access, geo-blocking)
3:56 — AI as shadow IT vs. sanctioned enterprise tools
5:38 — LLM observability, PII filtering & the agent visibility problem
7:09 — IT on the back foot: productivity vs. security trade-offs
8:07 — The VP of AI role and cross-company bridge-building
9:17 — Token burn, ROI quantification & compliance apps as a first step
11:36 — Meta’s token-burn employee evaluation: a cautionary tale
12:41 — Enablement, Claude Code, and structured agentic development
13:31 — Managing 20,000 virtual developers: same rules apply
13:45 — Developer reactions to AI: skeptics vs. power users
14:37 — QA as the persistent bottleneck in agentic coding
15:24 — Real-world AI spend blowouts: Uber’s 2026 budget story
16:16 — Is the token burn actually worth it? Innovation vs. ROI
17:59 — Chasing share price vs. genuine productivity gains
19:11 — Microsoft AI costs exceeding employee costs; compliance apps as guardrails
21:41 — Hiring junior devs back after over-automating; market overcorrections
22:43 — Future hiring tiers: agentically-enhanced senior devs & token budgets
24:14 — Building a good internal AI strategy: low-hanging fruit & handoff automation
26:26 — Proving ROI to boards; Mobot & custom dashboards in Sumo Logic
27:38 — AGI expectations vs. reality; creative thinkers as AI’s biggest winners
28:45 — Plugins, skills & defining what “success” looks like with AI
30:10 — Sanctioned AI use enabling real productivity (the spreadsheet example)
32:59 — Claude Code remote control, voice dictation & async agentic work
33:56 — claude.md context loading tip & persistent project context
35:07 — Wrap-up: navigating the ever-evolving digital landscape