How ‘tokenomics’ exposes AI’s true business case.

As the era of cheap, unlimited language models ends, Aiimi CTO Paul Maker explains why businesses must rethink their approach – focusing on cost, value, and sustainable AI architecture before chasing the latest technology. 

Since the advent of language models, access has largely been ‘all-you-can-eat’. Whether you were using coding tools like Claude or Cursor, or had a standard ChatGPT subscription, $20 a month bought you almost unlimited use, with most users rarely exceeding their daily quota. These services were heavily subsidised by venture capital, which made them artificially cheap. Now, as funding dries up and IPOs approach – Anthropic, for example, is targeting a valuation close to SpaceX – these companies need to show a clear path to profitability. The days of giving away the cake for free are over.

Rethinking AI ROI

In the past month, a wave of AI services and tools have shifted from pay-as-you-go subscriptions to pay-as-you-use or pay-per-token models. The price hikes have been dramatic; for some, costs have jumped from $20 a month to $1,000 or even $3,000. The cheapest models might cost one unit, but the most advanced models are significantly more expensive. The real-world costs are finally being felt, and there’s nothing on the horizon to suggest prices will come back down. The technology is fundamentally resource-hungry, demanding huge data centres, vast amounts of water, and enormous energy consumption. Unlike other technologies, where scaling up brings the unit cost down, the economics here simply do not improve with scale.  

It’s telling that organisations like Meta and Amazon were, until recently, encouraging employees to use as much generative AI as possible – what some called “tokenmaxxing”, measuring engineers by token usage rather than quality of work. Nvidia CEO Jensen Huang even said he’d be “deeply alarmed” if a $500,000 engineer did not burn at least $250,000 worth of tokens in a year. But almost overnight, these organisations have reversed course. This reflects a growing recognition that when applied to the wrong use case, AI can be more costly and less effective than humans. 

At the same time, 74% of businesses say AI adoption is now a top-three strategic priority. So, what’s to be done? ‘Tokenomics’ is about managing the use of language models in a sustainable way – not just throwing money at the most expensive models for every task, but weighing up whether there is a genuine return on investment and matching the right model to the right job. Language models are no longer a free lunch, so using them demands careful consideration.  

Building a sustainable AI strategy

Asking a question and receiving an answer in ChatGPT is a relatively cheap interaction. But for agentic tasks like coding, these platforms need to read huge amounts of context and can get caught in loops, talking to themselves to arrive at an answer. This can burn through millions of tokens, making the costs astronomical and far higher than paying a software developer to do the same task. Unlike humans, language models start from scratch every time – they don’t remember your codebase or previous interactions, so every prompt is a new cost. This highlights two key points: first, you need to be aware of token usage and whether it delivers any real ROI for that use case. If not, stop. Second, you need to consider whether you’re using the right model for the job – including whether the cost of agentic AI is valid for that task. 

This shift is driving a new kind of AI conversation: what policies and frameworks do you need to balance paying the right amount with getting the quality you need? Whether you call it AI architecture, AI strategy, or AI policy, the core principle remains the same: organisations need to develop a clear, deliberate approach to AI.  

AI outcomes over hype

It isn’t about chasing the latest model or following every new trend. It’s about understanding your business needs, making considered choices on where and how AI is used, and being honest about where AI adds value – and where it simply doesn’t justify the cost. This is exactly the work we’re doing with our customers: putting the tools and controls in place to match the right model to the right task and align AI use cases to real business outcomes. Too many organisations have been caught up in the hype, investing in AI initiatives that don’t deliver meaningful results. Now, with costs skyrocketing and scrutiny increasing, it’s more important than ever to focus on genuine business value, not technological novelty. 

The era of easy wins and unlimited experimentation is over. The organisations that will succeed with AI from here will be those that treat it as a strategic capability – managed, measured, and always aligned to real outcomes. That’s the real lesson of tokenomics: sustainable, value-driven AI isn’t just a technical challenge, it’s a business imperative. As the hype fades, the focus must now be on applying AI where it creates real advantage.

Learn more about our AI Strategy & Roadmap Services.