Key Takeaways

  • Torsten Slok at Apollo reports that Fortune 500 companies outside the Magnificent Seven tech giants are suffering from enterprise AI indigestion.
  • Goldman Sachs recently clamped down on non-technical staff and traders vibe coding their own software dashboards, citing dual concerns around operational risk and weak productivity returns.
  • Enterprise AI adoption is hitting the messy jobs thesis, an economic reality where most corporate roles consist of thirty fragmented tasks rather than one cleanly automatable workflow.
  • Engineering departments remain the only enterprise teams with the political capital to request a ten-thousand-dollar monthly AI budget.
  • The broader AI market trade requires an immediate jump in baseline model capabilities before standard corporate buyers will expand their software budgets.

The Fortune 500 Hits AI Indigestion

The Magnificent Seven tech giants keep pouring billions into compute, but the rest of the corporate world is tapping the brakes. On the floor of the New York Stock Exchange, John Coogan pointed to recent analysis from Apollo Chief Economist Torsten Slok. Slok found that Fortune 500 companies outside of Big Tech are experiencing quiet indigestion over their AI investments.

Boardrooms approved exploratory software budgets throughout the past two years, but finance departments now want to see actual margin improvements. They are not finding them. As Coogan noted, “right now there's no one outside of an engineer that can go to their boss and say I want to spend $10,000 a month on AI.” (play from 13:35) Unless software drives direct headcount reduction or billable revenue, enterprise buyers treat experimental AI tools as pure corporate overhead.

Goldman Sachs Blocks Internal Hackers

When consumer AI coding tools took off, tech evangelists promised that every corporate worker would build custom software. Business analysts and investment bankers would throw together bespoke dashboards before lunch, bypassing centralized IT queues. That experiment just hit Wall Street compliance walls.

Joe Weisenthal highlighted that Goldman Sachs recently stepped in to stop its trading desks and non-technical staff from writing their own ad-hoc software tools. “Goldman is clamping down on its non-technical employees like vibe coding, including its traders,” (play from 13:47) Weisenthal explained, “and it said it was for both risk and productivity.” (play from 13:57) A rogue dashboard running unverified Python code inside a trading desk introduces compliance exposure, while building throwaway tools pulls traders away from making markets.

The Messy Jobs Problem

Beyond compliance panic, AI tools face a structural adoption barrier inside traditional corporations. Weisenthal calls this dynamic the messy jobs thesis. Most corporate knowledge jobs look nothing like software engineering, where a worker writes code against a clear technical spec.

Instead, real white-collar roles are webs of context. “We all have like, you know, within a job we have like 30 tasks,” (play from 14:38) Weisenthal said. A risk officer or supply chain manager spends ten minutes reviewing an invoice, twenty minutes soothing an anxious client on Slack, fifteen minutes reconciling an ERP table, and thirty minutes in an informal meeting untangling an internal political dispute. Automating one slice of that day saves forty minutes, but it does not let a company cut a salary or eliminate a seat. Coogan argued that this dynamic caps enterprise value unless models make another giant leap: “for the overall trade to work, you have to have a massive capabilities jump like soon.” (play from 12:41)

What to Do With This

Audit your product pitch before your next enterprise sales call. If your software automates a single recurring task for an operations role, stop pitching labor replacement. Build your pricing and metrics around clean risk reduction or audit compliance, because finance teams will not approve five-figure monthly software bills on vague personal productivity gains.