Uber Burned Its Entire AI Budget in 4 Months. Here’s What Australian Small Businesses Should Do Instead

Uber told its engineers to use AI as much as possible. The company even ranked them on internal leaderboards by how much AI they used. The result: Uber burned through its entire 2026 AI coding budget in four months. Per-engineer costs hit $500–$2,000 a month on tools like Claude Code. Now the company has capped spending at $1,500 per employee per month, and its own COO can’t tell you whether any of it was worth it.

“It’s very hard to draw a line,” Uber COO Andrew Macdonald said on a podcast this month, between all that AI spending and “25% more useful consumer features.” Translation: they spent a year’s budget in a third of the year, and the ROI is anyone’s guess.

This came up at the Sydney Data & Analytics Summit this week, where Gartner researchers told the room that fewer than one in five companies are actually using AI well. That’s not a Sydney problem or a startup problem. It’s an industry-wide one.

The broader pattern is worse than Uber’s story

Uber’s not alone. Microsoft pulled its developers’ Claude Code licences after costs blew out. When Priceline went to renew its Cursor contract, the price came back four to five times higher. OpenAI’s own head of enterprise says every customer conversation has shifted from “what can AI do?” to “what token controls do you have?”

TechCrunch has started calling it “token-maxxing”: companies pushing AI usage to the max with no guardrails, then panicking when the bill arrives.

The numbers back it up. Gartner says 30% of generative AI projects get abandoned after the proof-of-concept stage. BCG research puts only 5% of companies in the “AI leader” category, with the other 95% seeing minimal returns or none at all. And 80-85% of enterprises miss their AI cost forecasts by more than 25%.

These are big-company stats. But the pattern shows up at every scale.

So what does this mean if you run a small business in Australia?

You’re probably already paying for at least one AI tool: ChatGPT, Copilot, an AI-assisted design or writing tool. Maybe more. And if you’re not actively tracking what each one costs you versus what it produces, you’re doing exactly what Uber did, just with fewer zeros.

Here are three rules worth applying now, before the budget surprise arrives.

Rule 1: Set a monthly spend cap before you start, not after.

Uber set caps after it burned through an annual budget in four months. Don’t wait for the crisis. Decide upfront what you’re willing to spend on AI tools each month across your whole business. Write it down. Review it monthly. If a tool can’t justify its line item within 60 days, cut it.

Rule 2: Measure output, not usage.

Uber ranked engineers by how much they used AI. This is exactly backwards. High usage with no measurable output isn’t productivity. It’s spend. For your business, the question isn’t “are my staff using AI?” It’s “what did AI help us produce this month that we couldn’t have done otherwise, and did that justify the cost?” Hours saved, jobs completed faster, content produced. Pick a metric that maps to your actual revenue, and track it.

Rule 3: Choose tools that fit your work, not tools engineers use.

Claude Code and Cursor are built for developers writing code all day. Most small businesses don’t need them. A tradie business, a boutique law firm, a café chain, a small ecommerce store. The right AI tool for you looks very different from what a Silicon Valley engineering team uses. Start with one tool that solves a specific problem you have right now. Not a general-purpose AI subscription. A specific solution.

The real lesson from Sydney

The Gartner finding that landed at the Sydney summit wasn’t a warning about AI being bad or overpriced. It was a warning about governance, specifically the lack of it. Companies that are getting real value from AI have two things in common: they started small, and they measured everything.

The businesses struggling are the ones that treated AI adoption like a culture war they had to win fast. Uber put its engineers in a competition to use more AI. That’s not a strategy, that’s how you blow a budget.

You don’t need to move slow. But you do need to know what you’re spending, what you’re getting, and when to stop. That’s not a constraint on AI. That’s just how you run a business.

Sources: TechCrunch, June 2; TechCrunch, June 5; Fortune, May 26; Gartner, July 2024

📖 More deep dives: News Deep Dives

📖 More deep dives: News Deep Dives

📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs

More hot takes: Hot Takes →