Procurement (rightfully) bares its fangs

The AI revolution isn't stopping. But the era of the blank check is over.

Author

Malcolm Frank, CEO
June 08, 2026

Procurement (rightfully) bares its fangs

Imagine letting a pack of ten-year-olds loose in a candy store, declaring everything free, and praising them for every piece swallowed. It would be a spectacular way to induce a medical emergency.

For the past year this was corporate America’s AI strategy, the tokenmaxxing binge, and the results were predictably chaotic. One company allegedly torched $500 million in a single month after launching Claude company-wide without usage caps.  In another case, an employee had burned through over $400 in tokens using AI to check the weather.  Such anecdotes abound, but all end up in the same place: with finance teams in a fury. 

How did the bill get this high? Because developers have been defaulting to the most sophisticated frontier models for virtually all their daily work. It is massive technical overkill, exposed by a jaw-dropping disparity in basic token economics.

Consider a standard enterprise team spinning up an agentic workflow consuming a modest 1 billion input and 1 billion output tokens a month. Depending on the logo on the model, that monthly invoice ranges from $210,000 down to a mere $1,300.

When the delta between the market leader and a fully-capable alternative is a staggering 160x price premium, the "infinite capability gap" defense completely crumbles. Paying a quarter-million dollars for a task a capable alternative can execute for the price of a nice team dinner isn't a tech strategy, it's fiduciary negligence.

Enter procurement, wielding a.) hard spending caps and b.) intelligent model routing software. These model routing orchestration solutions have instantly become the hottest thing in enterprise tech, acting as financial traffic cops for your AI consumption. The goal is simple: stop hiring a 40-ton dump truck to haul three pounds of sand. By triaging the routine 80% of queries to cheap, commoditized models and hoarding the expensive frontier engines for genuine complexity, organizations can slash their AI spend by 60% to 80% overnight.

Of course, this procurement crackdown will blunt the growth curves of the frontier model giants, raising uncomfortable questions about how much recent revenue was “real.” And with the IPOs of Anthropic and OpenAI looming, the street is suddenly scrambling to model realistic growth rates and paths to profitability.

The AI revolution isn't stopping. But the era of the blank check is over. Mom just walked into the candy store, and she's holding the receipts.

Model routing threatens OpenAI and Anthropic’s revenue model

The enterprise shift to model routing cuts AI costs but threatens the revenue growth of model providers.

Mystery company accidentally blew $500 million on Claude AI in a single month — failed to put usage limit on licenses for employees

AI sticker shock is rattling enterprise tech. The new bottom line: Cap your tokens—or your model will devour your cash flow.

AI’s Black Friday

Gary Marcus argues Friday’s selloff isn’t a routine correction. Instead, it’s a structural reckoning for a tech sector that poured billions into scaling language strings while ignoring the absolute absence of a path to profitability.

Anthropic urges AI labs to pause, warns humans risk losing control

Company warns that rapid advances in technology could soon allow AI systems to improve themselves faster than society can control the risks. Is it legit? Skeptics think the model leader is catastrophizing, angling for regulatory capture. 

Ethan Mollick: The new rules of living with AI agents

When AI handles the execution, human success depends entirely on knowing when to refuse the shortcut.

AI Is changing how consultants get paid—and much more, BCG’s CEO says

Christoph Schweizer points to higher revenues and head count in response to predictions the technology is killing his industry.

AI is rewriting the economics of outsourcing

The decades-old "labor arbitrage" model is collapsing. In the agentic era, companies will no longer pay for hours worked—they will pay for results achieved.

Satya Nadella says AI agents should be treated like employees with identities, permissions, and audits

The workplace of tomorrow is hybrid in a whole new way. Microsoft's blueprint for the agentic era means treating AI systems like members of the team, complete with corporate governance.

AI agents now generate more web traffic than humans do

Cloudflare CEO Matthew Prince says he didn't expect this milestone until 2027.

Solution of the week:
Regulatory & Policy Change Monitor

The Regulatory & Policy Change Monitor tracks regulatory updates relevant to your business continuously, compares changes against your current practices, and generates plain-language impact summaries with recommended actions.

Anthropic says 80% of its new production code is now authored by Claude – how your enterprise can keep up

With Claude driving an eightfold explosion in engineer output, the new technical imperative isn't writing lines of code—it’s mastering automated quality control and architectural oversight.

Kirkland & Ellis’ controversial $500 million AI bet

The law titan is building its own model. If this massive project works, it will widen the gap between elite firms and the rest of the legal system. If it fails, it will be a costly reminder that corporate cash cannot easily buy a proprietary AI edge.

“Organizations are layering AI on top of existing chaos”

AI is speeding up work for individual employees, but businesses-wide productivity is floundering.

Infosys, TCS, and Wipro now have 300,000+ employees using Copilot as AI moves into daily work

With Wipro alone reclaiming 250,000 workdays this quarter, this massive deployment proves generative AI has graduated from an experimental novelty into core operational infrastructure.

And one more thing…

This will be the first World Cup with AI coaches on the sidelines

It will help smaller teams more than bigger teams.

Reply

or to participate.