The AI Budget Hangover Is Real — and Enterprises Are Still Searching for ROI

Earlier this year, Silicon Valley was gripped by a phenomenon called tokenmaxxing — a strategy where executives actively pushed employees to use AI as aggressively and as frequently as possible. The logic seemed straightforward: more usage equals more value. Then the invoices arrived.

According to a report by Theresa Loconsolo at TechCrunch AI, the consequences were swift and painful. Uber reportedly burned through its entire annual AI budget in just a few months. Some companies began cutting Claude licenses for entire departments. Meta quietly killed its internal AI usage leaderboard. The message from enterprise boardrooms is now increasingly clear: unchecked AI consumption is not a strategy.

NEA partner Tiffany Luck, speaking on the matter, says enterprises are still in the process of figuring out what AI actually returns on investment — and that tension between enthusiasm and accountability is defining how companies approach AI spending right now.

What Tokenmaxxing Actually Revealed

The tokenmaxxing trend was not without logic at its core. If AI tools can accelerate work, then getting teams to use them more should, in theory, produce more value. The problem is that raw usage volume was treated as a proxy for business impact — and those are two very different things.

When Uber blew past its AI budget in record time, it was not because its teams were generating proportionally more revenue or shipping faster. It was because consumption and output had been decoupled. AI spend scaled up; measurable business results did not keep pace. That is a fundamental ROI problem, and it is one that even the most well-resourced enterprises are struggling to solve.

The decision by some organizations to pull Claude licenses from portions of their workforce is a telling signal. It suggests that access was being granted broadly, without clear use-case discipline. When the budget ceiling arrived, the cuts were indiscriminate — not strategic.

Why This Matters Beyond the Enterprise World

It would be easy to read this as a story about large-company excess — a cautionary tale relevant only to organizations with massive AI budgets. That would be a mistake.

The same dynamic plays out at smaller scale for SMBs and mid-market teams. Subscription creep is real. A business might be paying for five or six AI tools simultaneously, with overlapping capabilities and no clear owner for any of them. Without a framework for measuring what each tool contributes — in time saved, revenue influenced, or errors reduced — you are essentially running your own tokenmaxxing experiment, just with smaller numbers.

The lesson from the enterprise reckoning is not that AI tools are overpriced or overhyped. It is that deployment without accountability produces waste at every company size.

Building an AI ROI Framework That Actually Works

So what does a disciplined approach look like? A few principles are emerging from companies that are getting this right.

Tie AI usage to specific workflows. Rather than giving broad access to a tool and hoping teams find value, assign AI tools to defined processes — customer support triage, first-draft content generation, data summarization — and measure the output against a baseline.

Track cost per outcome, not cost per seat. Seat-based pricing models encourage license hoarding. Outcome-based thinking forces you to ask whether the AI investment is actually producing something measurable.

Audit regularly. The companies that cut Claude licenses did so reactively, after spending had already exceeded projections. Proactive monthly or quarterly audits of AI tool ROI prevent that kind of forced, disruptive pullback.

Consolidate where possible. Many teams are using separate tools for writing assistance, research, scheduling, and workflow automation when a single AI productivity platform could cover the majority of those use cases at lower total cost.

For teams trying to build that kind of AI strategy for SMBs, the goal is not to use AI less — it is to use it with more intention.

The Bigger Picture

Tiffany Luck's observation that enterprises are still figuring out their AI ROI is not a pessimistic verdict. It is an honest description of where the market is. The tools are genuinely powerful. The business cases are real. But the measurement infrastructure — the layer of accountability that connects AI spend to business outcomes — is still being built across most organizations.

That gap is exactly where platforms like WRRK.ai are positioned to help, giving teams a structured way to deploy AI across their operations without losing visibility into what it is actually delivering.

The tokenmaxxing era taught us that enthusiasm is not a strategy. The companies that come out ahead will be the ones that pair adoption with accountability.

Original reporting by Theresa Loconsolo, TechCrunch AI. Published June 17, 2026. Read the original here.

Ready to deploy AI in your business without the budget surprises? See how WRRK.ai helps teams measure what matters — visit WRRK.ai.

Frequently Asked Questions

What is tokenmaxxing and why did it backfire for enterprises?

Tokenmaxxing refers to the practice of encouraging employees to maximize their AI tool usage as broadly and frequently as possible. It backfired because organizations treated consumption volume as a measure of value, rather than tracking whether that usage was producing measurable business outcomes. When budgets ran out — as reportedly happened at Uber — companies were left with high costs and unclear returns.

How do you measure ROI on AI tools for your business?

Measuring AI ROI starts with tying each tool to a specific workflow and establishing a baseline for that process before AI is introduced. From there, track cost per outcome — such as time saved per task, reduction in error rates, or revenue influenced — rather than simply counting seats or logins. Regular audits help catch underperforming tools before they become budget problems.

Should small businesses be worried about AI budget overruns?

Yes, though the scale is different. SMBs face the same risk of subscription creep and undisciplined AI tool deployment as large enterprises, just with tighter margins for error. The same principles apply: assign tools to specific workflows, measure results against a baseline, and consolidate overlapping capabilities wherever possible to keep costs aligned with outcomes.

The AI Budget Hangover Is Real — and Enterprises Are Still Searching for ROI