Enterprises are pouring billions into artificial intelligence, yet 85% of companies admit they have no clear method to measure if these investments are actually yielding results. This stark reality formed the crux of a recent a16z podcast featuring Russell Fradin, CEO and Co-founder of Larridin, and Alex Rampell, General Partner at a16z. Fradin, a seasoned entrepreneur with a history in ad tech, argued that the AI industry is currently grappling with a fundamental measurement and attribution problem strikingly similar to the early days of internet advertising.
Fradin, who sold his first company for $300 million and was a foundational executive at comScore, spoke with Rampell about the critical missing infrastructure in AI adoption. Their conversation highlighted why the measurement systems that fueled the trillion-dollar boom in online advertising are precisely what AI needs to move beyond its current state of unquantified potential. The core issue, as articulated by Fradin, is that companies are buying AI tools without understanding whether anyone is truly using them, let alone if they are driving actual productivity gains.
A central insight of the discussion was the parallel drawn between the nascent AI market and the early internet advertising landscape. In the late 1990s, advertisers invested heavily in banner ads without clear metrics for their effectiveness. "A lot of ad tech is here's an advertisement, and there's this attribution problem," Rampell noted, questioning, "Who is responsible for that sale?" This lack of attribution and measurement infrastructure stifled growth until companies like comScore provided verifiable data. Fradin contends AI faces the same hurdle: "The technology is unbelievable... but also there are very boring but important questions." These questions revolve around whether AI tools deliver tangible benefits or merely inflate costs.
The current state of AI adoption within large enterprises is often characterized by a "hope-and-pray" strategy. Companies purchase AI solutions, but lack visibility into actual employee usage or the impact on workflows. Fradin shared a telling anecdote: a 28-year-old investment banking analyst used ChatGPT to create a 30-slide deck in minutes, a task that would typically take eight hours. Instead of recognizing this individual's innovative use, the company organized a global call to teach everyone how to use ChatGPT. This, Fradin stressed, is "an absurd way to hope people adopt world-changing technology."
Another critical insight is the behavioral aspect of AI adoption. Many productive employees are reportedly "hiding their AI usage from management." This clandestine adoption stems from a fear of looking "dumb" or, more profoundly, anxiety about job security. Companies are not providing the necessary training or safe spaces for employees to experiment and integrate AI tools openly. This creates a disconnect where management invests in tools, but widespread, transparent, and measured adoption remains elusive.
The conversation also touched upon Goodhart's Law, which Rampell invoked: "When a measure becomes a target, it is no longer accurate as a measure." This principle underscores the danger of relying on easily quantifiable but ultimately misleading metrics. For instance, simply tracking the number of AI tools purchased or the lines of code generated by an AI assistant might not reflect genuine productivity or value. Instead, enterprises require sophisticated measurement that ties AI usage to actual outcomes and business objectives, not just input metrics. Without robust, independent measurement, companies might optimize for the wrong things, creating an illusion of progress.
Fradin emphasized that Larridin’s mission is to address this very problem by providing comprehensive measurement and governance tools for enterprise AI. Their approach begins with understanding what tools are actually present in a company and if employees are using them. This behavioral data is then married with traditional productivity research, creating a robust baseline. The ultimate goal is to move beyond mere usage tracking to determine if AI is making employees "fundamentally more productive." This shift in focus from inputs to outcomes is essential for unlocking the full potential of AI.
Related Reading
- The Unmeasured Billions: Unlocking AI's Hidden Productivity
- AI's Enduring Rally: Navigating Growth and Diversification in a Broadening Market
The economic implications are significant. Rampell highlighted how AI represents a shift where "software is eating labor." While this doesn't necessarily mean job loss, it implies that existing workforces will become significantly more productive, potentially ten times over. This transformation will lead to a dramatic reallocation of budgets from labor to software. As companies increase their software spend, particularly on AI, the demand for clear ROI will intensify. "The numbers are just big," Fradin noted, pointing out that AI spending is "going way beyond experimental."
Larridin aims to be the "best friend to all of the AI companies" by enabling this crucial measurement. They are building the infrastructure for measurement and governance, not to stifle innovation, but to accelerate it. By providing clarity on AI's impact, Larridin seeks to empower companies to strategically deploy and scale their AI investments, ensuring that the billions being spent translate into measurable value.
