As artificial intelligence (AI) reimagines nearly every longstanding process in recent years, the technology landscape has been tumultuous, to say the least.
Globally, teams are being expected to respond to the sudden disruption, drive innovation, and navigate market shifts, all while proving ROI (return on investment) and juggling competing priorities. Sounds challenging, right? In fact, companies are so desperate to jump onto the AI bandwagon that they’ve been pushing employees to use AI as much as possible, even if they aren’t able to entirely understand what it is, what it can do, and how exactly it could be helpful.
Adding fuel to the fire is the concept of tokenmaxxing, where tech firms are incentivising employees based on how much AI they use through their usage of AI tokens. However, tokenmaxxing is facing something of a confidence crisis, and a new trend is emerging that focuses on more intentional AI investment rather than just burning through AI tokens – and it’s aptly called ‘Valuemaxxing.’
The Financial Argument Of Tokenmaxxing
Breaking onto the tech scene in early 2026, tokenmaxxing involved businesses encouraging software developers to consume as many AI tokens as possible to drive innovation. Unsurprisingly, the practice turned using token consumption as an imperfect productivity metric, and it wasn’t just some fringe trend.
For instance, Nvidia’s CEO commented that he’d be “deeply alarmed” if a USD 500,000 developer spent less than USD 250,000 on AI tokens! Not just that, but token leaderboards, which were pioneered by Meta engineers, found their way to even the likes of Disney.
Despite its impressive momentum, not only did the controversial practice throw open the doors to catastrophic expenditure, but also it turned into something of a vanity metric, and is, thus, facing a confidence crisis. In May 2026, Microsoft announced that it would be cutting a majority of its Claude Code subscriptions amidst financial concerns.
Likewise, Uber had burnt through its entire AI budget in just four months. If that wasn’t enough, a story broke of an AI consultant who claimed that a client spent USD 500 million in a month due to no one having put any usage limits on Claude licenses for employees.
All this goes to show that despite its impressive momentum, spending tokens doesn’t feed any direct value to the business nor does it have any tangible business impact. Tokens might be a great measure of input, but they aren’t a great measure of outcomes or output. This lack of connection, especially given the high cost of compute, is what makes tokenmaxxing very difficult to justify.
Valuemaxxing: Putting Judgment Back in Charge
With tokenmaxxing appearing to subside, other approaches are emerging in its place, and valuemaxxing has particularly stood out as a critical market-wide reset on how businesses manage technology risk and spend in an AI-driven world. While tokenmaxxing focuses on token usage, valuemaxxing moves the conversation toward direct outcomes by challenging the need for more spend, tokens, or tools. Thus, it pushes teams for more details on control, accountability, and return, shifting the focus to making more intentional, smarter investments that drive real ROI rather than spending more on technology.
So, what makes valuemaxxing relevant? Valuemaxxing begins where counting tokens ends. Let’s be honest: the first phase of GenAI saw companies tracking usage dashboards and teams measuring prompts, a phase that was necessary at that time. However, despite AI being firmly in use (by 2024, 75% of knowledge workers were using GenAI at work), high token counts shouldn’t be policy goals as it not only conceals shadow costs but also conflates input with output.
Valuemaxxing, on the other hand, leverages useful insight derived from tokenmaxxing, rejecting the superficial measure.
Another aspect that valuemaxxing brings to the table is context. In an enterprise study with AI assistance, customer service agents could solve almost 14% more queries every hour. Meanwhile, a writing performance study found that participants finished not only faster but also exhibiting better quality.
However, the study also showcased an important pitfall: once the scope of work was not within AI’s reach, users employing AI weren’t always able to find the right answer. So, AI doesn’t have a good or bad quality; rather, its value is moderated by the context – the human oversight and the suitability of the task.
AI Governance By Value Rather Than Volume
Valuemaxxing needs to be adopted as a straightforward rule, defined around human tasks. While some require value in quality – novelty and innovation and accuracy and data validation – others require value in a particular direction – knowledge-critical decisions and speed, such as summarising routine data.
Depending on the type, the minimum rules for AI need to be different, as any one-size-fits-all policy will be either too constrained or too free – and valuemaxxing is a better standard in every sense of the word.
It’s clear that rolling out AI isn’t necessarily a silver bullet, and providing employees with access to AI isn’t going to make them more productive by default. Getting value from AI comes down to setting up more structured and stronger oversight, avoiding shortcuts, demanding visibility into tech usage, and eliminating redundant applications.
In case you missed:
- Think You’ve Seen Everything in AI? Welcome To The Controversy That Is ‘Tokenmaxxing’
- Are AI Tokens the New Digital Currency?
- Hiding In The Dark: Navigating The Threat Of Shadow AI
- NFTs vs. RWA Tokenization – The Future of Digital Asset Ownership
- Pi Day Update: What’s Happening With The Controversial Cryptocurrency?
- Tokenization of Real-World Assets: Assets Going Digital?
- Agentic AI Is Now A Fund Manager, Believe It Or Not
- Is AI Changing The Rules For SaaS Pricing Models?
- The Trust Deficit in AI – What You Need To Know
- How Blockchain and AI Are Revolutionizing Modern Finance
