For more than twenty-five years, I have watched technology cycles oscillate between the “magic” phase and the “utility” phase. I remember the early days at Nokia when we treated the mobile phone as a portable telephone, only to later realize it was actually a remote control for life. We are currently at a similar pivot point with Artificial Intelligence, but there is a growing gap between the “clever hack” and the sustainable product. If we are not careful, we aren’t just building agents; we are automating the next wave of enshittification.
The honeymoon phase of the AI revolution is officially over. Boards and leadership teams are no longer satisfied with flashy proofs-of-concept or agents that “experiment” in safe, isolated playgrounds. We have entered the era of the reckoning: Can your organization actually run AI reliably, responsibly, and repeatedly inside real business processes? As highlighted in recent industry analysis, we are shifting from raw model intelligence to predictable agentic control.
The Trap of the “Rot Economy”
In our rush to automate, we must be wary of what Cory Doctorow famously termed “enshittification”—the process where platforms eventually cannibalize the user experience to extract maximum value. We are starting to see the symptoms in the “data tolls” and increasing connector fees from major SaaS players. When technology stops improving the user experience and starts serving the corporate bureaucracy, innovation dies.
For product leaders, the challenge is protecting the user from this “Rot Economy.” We see this at companies like Salesforce, which has faced scrutiny over shifting fee structures for data access. If your AI strategy relies on third-party APIs that can degrade in quality or spike in price without warning, you aren’t building a product; you’re building a dependency. This is why many pragmatic leaders are now pivoting toward Small Language Models (SLMs) that they can own and govern internally.
Moving Beyond the Chatbox: The Agentic Era
The era of Generative AI as a fancy writing assistant is ending. We are entering the Agentic Era. But as with the .com boom and the mobile revolution, there is a fundamental difference between an “answer engine” and a product that creates value. According to the latest Reuters Institute Digital News Report, publishers are bracing for a 40% decline in search referrals because aggregators are consuming content without passing on the traffic.
The solution isn’t to fight the agents; it’s to build them. However, we must move beyond the “chat” interface. Consider the difference between a bot that tells you about flight prices and an agent that has the autonomy to book the flight, negotiate a refund when it’s delayed, and rebook you on a rival carrier based on your personal preferences. This requires a fundamental shift in our Product Ways of Working. You cannot build autonomous agents with siloed, project-based teams. You need empowered, cross-functional “product trios” that understand the nuances of the user journey.
The Small Model Advantage and Spatial Reasoning
While headlines focus on the giants like OpenAI, the real winners in the enterprise space are becoming the Small Language Models. Why? Because value for the user is found in speed, privacy, and cost-efficiency. Major players like AT&T are increasingly moving away from massive, out-of-the-box LLMs in favour of fine-tuned SLMs to maintain better guardrails against the rot economy.
Furthermore, we are hitting a wall with “brute-force” scaling. The next leap isn’t coming from a larger version of what we already have; it’s coming from “world models” and systems that understand spatial reasoning. Startups like World Labs and Runway are shifting the focus toward models that learn through physics and interaction rather than just predicting the next word in a sentence.
- Own your infrastructure: Avoid the “data tolls” by investing in models you can control.
- Focus on Outcomes: Agents should be measured by their ability to complete tasks, not the cleverness of their conversation.
- Empathy First: Technology cycles prove that the most “invisible” infrastructure wins. Design for the user, not the hype.
As we navigate this “Great Filtration,” the labs and businesses that simply throw more GPUs at the problem will be separated from those innovating on architecture and user value. Stop waiting for a smarter model to solve your product’s structural flaws. The “quiet transformation” happens when the technology becomes the invisible support for a better human experience. It is time to step out of the playground and start building for the real world.
Leave a Reply