AI is Becoming ‘Corporate Theater’: Domo CDO Warns Against the Danger of Tokenmaxxing
Table of Contents
The Anxiety of the AI Arms Race
For many executives and employees, the current state of artificial intelligence feels less like a technological breakthrough and more like a ticking clock. In a recent candid discussion, Chris Willis, the Chief Design Officer and futurist at data platform company Domo, questioned why there isn’t more widespread resentment toward the companies driving this frenzy.
During a recent trip to San Francisco—the epicenter of the AI boom and home to giants like OpenAI and Anthropic—Willis observed a disturbing trend: a pervasive sense of professional anxiety fueled by the aggressive marketing of LLMs. According to Willis, the industry has effectively weaponized fear, creating a narrative where the only options are to adopt immediately or face obsolescence.
“Fear is not a durable strategy for innovating,” Willis argues. He suggests that the pressure emanating from the C-suite has created an environment where careers feel precarious, regardless of an individual’s actual productivity or the technology’s actual utility.
The ‘Product Without a Spec’
At the heart of the chaos, Willis identifies a fundamental product failure. In traditional software development, a product is built against a specification—a clear set of goals defining who the user is and what the tool is designed to achieve. LLMs, however, have entered the market as a “product without a spec.”
The current value proposition for these models is essentially that they can do anything for anyone, in any language, at any time. While this versatility is technically impressive, Willis contends it is a nightmare for organizational leadership. Without a specific use case, companies are spending millions on AI tools and expecting innovation to spontaneously occur, which is a fundamental misunderstanding of how business evolution works.
This gap between capability and strategy has led to what Willis describes as “AI theater.” Companies are no longer innovating to solve problems; they are innovating to show the market and their shareholders that they are “doing something” with AI.
The Rise of Tokenmaxxing
One of the most visible symptoms of this theater is a phenomenon Willis calls “tokenmaxxing.” This occurs when organizations purchase massive amounts of access to AI models and pressure employees to use them as much as possible, measuring success by the volume of tokens processed rather than the value generated.
While an individual employee might feel more productive using a chatbot to draft emails or summarize meetings, Willis warns that this rarely translates to the company’s bottom line. When the metric for success is usage rather than outcome, the result is a series of expensive proof-of-concept projects that lack the durability, trust, and scalability required for actual deployment.
Returning to Human Judgment
The danger of ignoring the “human in the loop” is not theoretical. Willis points to the experience of Swedish fintech firm Klarna, which famously touted the replacement of customer service staff with AI, only to eventually find that human intervention was still necessary. “Frankly, no customer ever just wants to talk to your chatbot,” he notes.
To avoid these pitfalls, Willis suggests a “slow-mo” approach: starting with boring, high-utility automations. Instead of chasing moonshot goals, he advocates for solving simple, tangible problems—such as an app that audits company invoices for anomalies—where human judgment is used to verify the AI’s output rather than be replaced by it.
As budgets tighten and the initial euphoria of the generative AI wave fades, Willis predicts a looming corporate reckoning. He expects CFOs to eventually stop asking how much AI they can buy and start asking why the massive spend hasn’t resulted in measurable gains.
