An economist just told corporate America its AI bet isn't paying off
Issue #066 · 4 min read
An economist just told corporate America its AI bet isn't paying off
Zero financial payoff outside tech, plus OpenAI lands on Google Cloud and a wave of GitHub Copilot updates.
The big story
Here's a sentence that should stop you mid-scroll.
AI has delivered "zero financial payoff" to corporate America outside of tech, according to Apollo's chief economist — this is TechCrunch's write-up of the claim.
Not "mixed results." Not "too early to tell." Zero.
Think about what that means. Companies outside tech have spent two years buying licenses, running pilots, hiring "AI leads," and rewriting job descriptions to mention Copilot.
And the economist's read is that none of it shows up on the bottom line yet.
This lands the same week OpenAI is publishing case studies about Uber rolling ChatGPT Work out to 30,000 employees and AT&T rebuilding its legal intake around the same tools.
Those are real deployments, not pilots. So either the payoff is coming and just hasn't landed in the numbers, or a lot of companies are about to discover that adoption and return are two very different charts.
If you're the PM asked to justify next year's AI line item, this is the number your CFO already half-believes. Bring your own, or expect to lose the argument to this one.
What shipped
OpenAI's models are moving onto Google Cloud. Frontier models and Codex workflows are headed to Gemini's Model Garden, available through Google Cloud Marketplace — details here.
If your company standardized on Google Cloud for procurement reasons and has been stuck choosing between "the good models" and "the cloud we're allowed to buy from," that tension just eased. Worth flagging to whoever owns your vendor list.
GitHub shipped a cluster of Copilot and security updates worth a skim, not a deep read. The Copilot usage metrics API now reports on more third-party tools like OpenCode, and separately now breaks out custom skills, agents, and MCP servers by name instead of lumping them together.
If you've ever tried to tell your CFO exactly what your engineers are spending Copilot tokens on, this is the report you've been missing. Pair it with GitHub's new Copilot app for Teams, which turns channel notifications directly into code changes.
Linear's Loops can now be triggered from Slack and GitHub, not just from inside Linear. The changelog says events in those tools can now kick off automated product work without someone noticing an update and acting on it manually.
Small thing, but it's the difference between automation that watches your tools and automation that actually sits inside them. If your team already lives in Slack threads for triage, this closes a real gap.
What I'd actually do this week
- Pull your own AI spend-to-output numbers before someone hands you Apollo's. If you can't produce a real ROI figure for your team's AI tools by Friday, that's the actual finding — not the economist's.
- If you're on Google Cloud and have been running a shadow OpenAI contract to get around procurement, check whether the Model Garden listing changes that calculus.
- Turn on GitHub's breakdown by custom skill and MCP server for one team this week. You'll likely find three tools nobody remembers approving.
Reply and tell me what your own AI ROI number actually looks like — I'd genuinely like to know if anyone's beating zero.