Users Will Bring Their Own AI Agents, Not Embed Them

The current AI gold rush is a masterclass in second-order thinking, a discipline few companies truly grasp. While many rush to integrate AI features, often for the sake of appearing innovative, the real advantage lies not in the immediate functionality but in understanding how AI will fundamentally alter user behavior and product ecosystems. This conversation with Jason Fried reveals that the most potent AI applications aren't about replicating human tasks, but about enabling new workflows and leveraging external AI agents. The hidden consequence for businesses is that building custom AI features today might be a costly distraction from the inevitable future where users bring their own sophisticated AI agents, fundamentally changing how products are used and how competitive moats are built. Those who understand this shift will gain a significant advantage by focusing on product simplicity and agent compatibility, rather than chasing every fleeting AI trend. This analysis is crucial for product managers, engineers, and business leaders seeking to navigate the AI landscape beyond the hype.

The Unseen Architecture: Why Users Will Bring Their Own AI Agents

The prevailing narrative around AI in software development is one of integration: adding AI features directly into existing products. Jason Fried, however, points to a more profound, less obvious shift: the rise of external AI agents that users will bring to products. This isn't just a nuance; it's a fundamental re-architecting of how AI interacts with software, with significant implications for competitive advantage.

Many companies are currently investing heavily in building custom AI functionalities, a process Fried suggests is akin to building a house before understanding the neighborhood. The immediate payoff is the appearance of innovation. But the downstream effect is a potential race to build features that will soon be rendered redundant.

"Things have actually changed so much in the past month, like with Open Claw, for example, and 24/7 running agents. Things are just perpetually running for you, and the ability for agents just to log in as normal people, that people are going to end up bringing their own agents to our products and just have them be normal users."

This quote highlights the core of the argument. Instead of embedding AI, the future likely involves products becoming platforms for sophisticated, external AI agents. These agents, already capable of learning and operating autonomously, will log in as users, bringing their own knowledge and capabilities. The implication is that products that are simple, clear, and provide robust APIs or command-line interfaces will be best positioned to accommodate this influx of external intelligence. The "competitive advantage" here isn't in having the most advanced built-in AI, but in being the most accessible and compatible platform for these external agents. This requires a different kind of investment: not in feature development, but in architectural openness and user experience simplicity.

The Illusion of Immediate Productivity

Fried's personal use of AI tools offers a glimpse into this more nuanced application. He describes using AI not to replace his thinking, but to amplify his efficiency in specific, high-leverage tasks. His use of AI as an editor, for instance, is not about generating content, but about refining language to better match customer vernacular. This is a subtle but critical distinction: AI as a tool for understanding and connecting, rather than for producing.

Similarly, his use of AI for rapid prototyping--quickly mocking up UI elements or generating realistic sample data--illustrates how AI can accelerate the decision-making process. The benefit isn't just speed; it's the ability to quickly iterate on ideas without burdening colleagues. This avoids the "owner's word weighs a ton" problem, where a request from leadership can derail other important work, even with disclaimers.

"And for me, it's not so much like I can do things I couldn't do before, because there are certain things I could do before that I just chose not to do. So it's mostly like this is a huge speed-up of time, and I don't need to bother somebody else."

This perspective reveals a key consequence: AI can reclaim time that would otherwise be spent on tedious tasks or on requesting help from others. This reclaimed time can then be reinvested in more strategic, higher-value activities. The conventional wisdom might be to use AI to automate tasks and reduce headcount. Fried's approach suggests a more powerful outcome: using AI to empower existing individuals to do more, faster, and with less friction, thereby increasing overall team effectiveness without necessarily expanding the team size. The delayed payoff here is a more agile and efficient organization, capable of faster innovation cycles because individuals can self-serve on many tasks.

The Pitfall of "Shitty AI"

The polarization in customer feedback--some demanding AI features, others fearing "shitty AI" ruining their experience--underscores a critical challenge. Many companies are slapping AI onto existing features without a clear understanding of its value, leading to frustrating user experiences. Fried’s stance is to avoid this "AI first" approach, emphasizing that AI should be available but not intrusive.

"Our approach is always to be as straightforward as possible, as no-nonsense as possible. So like slathering AI on everything everywhere all the time is not going to be our approach. And I've used tools that are like that now where everything is like AI first, and I just think it's a bit of a novelty at the moment, and I think it's wearing thin in some ways."

This is where the true competitive advantage can be forged. By resisting the urge to integrate AI everywhere, 37signals can maintain product simplicity and focus on core functionality. The consequence of this deliberate restraint is a product that remains accessible and understandable to a broad user base, while still offering avenues for power users and external agents. The "delayed payoff" is the preservation of a clean, user-friendly product that doesn't alienate its existing customer base, while simultaneously laying the groundwork for future integration with external agents. This strategy avoids the immediate cost of building potentially obsolete AI features and instead focuses on building a robust, adaptable platform.

The Human Element as a Lasting Moat

In the realm of customer support, the temptation to offload to AI is strong, driven by perceived cost savings. However, Fried argues that highly trained, long-term human support staff represent a massive competitive advantage. While AI can handle simple queries, the nuanced understanding and empathetic connection provided by humans are irreplaceable, especially for complex issues or when customers are frustrated.

"I think humans are hugely important. I think we have a massive advantage because we have incredibly good humans on our support team, many of which have been here for many, many years... And I think it's a massive competitive advantage for us, and I would not want to give that up."

The consequence of prioritizing human support is a deeper customer relationship and a more resilient brand. While competitors might be cutting costs with AI-only support, 37signals is building loyalty and trust. This is a second-order positive effect: the immediate investment in experienced human support pays off in long-term customer retention and brand reputation, a moat that AI alone cannot easily replicate. The discomfort of not fully automating support is outweighed by the advantage of superior customer care and a more human-centric product experience.

Key Action Items

• Prioritize Product Simplicity and API Access: Over the next quarter, focus on ensuring Basecamp's core features are exceptionally clear and that its command-line interface (CLI) and APIs are robust. This lays the groundwork for seamless integration with future external AI agents.
• Develop an "Agent Compatibility" Framework: Within the next 6-12 months, begin defining and documenting how external AI agents can interact with 37signals products. This isn't about building AI, but about enabling others to bring their AI to your platform.
• Invest in Human Support Expertise: Continue to treat customer support as a strategic advantage. Over the next 1-2 years, focus on retaining and training your support team, ensuring deep product knowledge and strong customer empathy. This pays off in customer loyalty and brand differentiation.
• Resist "AI Everywhere" Temptation: For the next 18-24 months, consciously avoid integrating AI features for the sake of novelty. Instead, identify truly high-value, non-intrusive AI applications that align with product simplicity and user control.
• Empower Internal AI Use for Efficiency: Continue encouraging internal use of AI tools for tasks like editing and rapid prototyping, as described by Jason Fried. This provides immediate productivity gains without compromising core product strategy. This pays off in the short-to-medium term (quarters).
• Monitor External Agent Landscape: Dedicate ongoing resources (e.g., one engineer part-time) to track the evolution of AI agents and platforms (like Open Claw) over the next 1-2 years. This vigilance will inform future strategic decisions.
• Offer Both AI and Human Support Options: Within the next 3-6 months, ensure that while AI may offer quick answers for simple queries, a clear and accessible path to human support is always available. This balances efficiency with customer experience and avoids the "AI loop of frustration."