Building Software vs Getting Adopted: The New Barrier

For the first time, building software is no longer limited to those who can code. AI is removing the barrier between ideas and execution. The real advantage is shifting to those who understand the business, the customer, and the problem. Software is no longer defined by how it’s built — but by how well it reflects what actually needs to be built.

So now what? Can everyone build their own software with AI? Today, more people can build their own tools, test ideas faster, and validate businesses without relying on a full development team. And that’s incredible. We’ve been using this approach ourselves, and it works. Quick builds, when used well, are powerful. They allow you to answer the most important question early, whether the market actually cares. But there’s a point where that same speed starts working against you. I like to think of it as carpentry. Everyone looks at a chair and thinks they could build one. And yes, they probably can. But building a chair that doesn’t warp, doesn’t wobble, is ergonomic, and holds up over time requires structure. Software is no different. When the foundation isn’t designed for real usage, the system starts behaving in ways no one expected. Costs become unpredictable, flows become harder to control, and simple actions start creating disproportionate impact. At that point, the problem is no longer how fast it was built. It’s what it was built on. AI accelerates creation, but it doesn’t replace architecture. When the foundation is well structured, speed becomes an advantage. The system moves from something experimental to something that operates with control and is ready to grow without friction.

AI Agents: Moving Enterprise Software Beyond the User Interface I have been learning about AI agents and has been simplifying how to integrate for Claims handling, For decades, enterprise systems have been built around one core concept: the user interface. Claims handlers navigate multiple screens, systems and workflows to move a claim from First Notice of Loss to settlement. The software guides the process, but the human performs the work. AI agents introduce a fundamentally different model, now it’s all about how the application self learns, adopt and provide more value addition

If software gets easier and cheaper to produce, does quality become a bigger differentiator? The more I read about AI and software delivery, the more I think the bigger shift is not just speed, but volume. If software becomes easier and cheaper to produce, we are likely to get a lot more of it. From code and features to products, and updates. I do not think that automatically gives teams an advantage. In some cases, it may just create more noise, rework, coordination overhead, and more risk. So if software gets easier to produce, does quality become a bigger differentiator? 🤔 Not just in the narrow sense of defects, but in the broader sense of trust. Things like: - Can teams build in enough quality that they are not drowning in verification and rework? - Can other teams use what you produce with confidence, rather than needing to recheck everything themselves? - Can you increase output without degrading the overall health of the system? That feels like a more useful question to me than whether AI makes teams faster. I wrote more about that in my latest Linky, including thoughts on trust between teams, release and post-release quality, and why more software does not automatically mean more value. 🔗 Link in comments. ❓What do you think becomes more important when software gets cheaper to produce?

Most enterprise software treats every user the same. Same interface. Same workflow. Same experience. 🥱 Tom Chavez, co-founder of super{set}, shares that agentic AI breaks that mold and companies that figure it out won't just have better software, they'll have a compounding advantage the competition can't catch up to. Tom calls it the Personalization Flywheel. Once it starts spinning, it’s hard to stop and even harder to compete with. The question is: what actually makes it work? Link in comments 👇

AI has forever changed the software landscape and made it possible for individuals with no knowledge of code to create their own software to meet their needs. Understanding what's possible versus what's sensible when making decisions on what to build versus buy is crucial to determining your success.

AI has forever changed the software landscape and made it possible for individuals with no knowledge of code to create their own software to meet their needs. Understanding what's possible versus what's sensible when making decisions on what to build versus buy is crucial to determining your success.

software is getting cheaper. fast. same product. same functionality. 90% price drop. and now with AI this will accelerate even more. building is no longer the hard part. distribution is. data is. integration is. features don’t protect you anymore. someone can rebuild it in weeks. what matters now: where you sit in the workflow. how deep you are embedded. how hard it is to replace you. this is what I’m seeing across products today. AI is not just improving software. it is commoditizing it. the question is simple: if your product disappears tomorrow, what stays?

AI has fundamentally changed how quickly software can be imagined and validated. However, the gap between working prototype and production-ready system remains a critical challenge. Organizations that successfully bridge this gap, combining AI-driven speed with engineering rigor, will be best positioned to build scalable, secure, and adaptable software platforms. The future of software development is not just faster ideation, it is disciplined execution at scale.

What happens to software when we design for the agent as the user and not a human? One of the biggest jobs a product manager has to accomplish is to curate what does and does not make it into a product. For the most part, this is a reflection on us humans' capability to absorb complexity and our need to have software feel like it was designed "for us". One of the biggest bottlenecks of product work is designing interfaces that feel consistent and elegant and user friendly for all of these features. Agents don't care about either of those things. At all. If you have 5 features or 5000 features, it's immaterial to an agent. Agents don't care if your labels have perfect spacing between them and the input box they represent. What's not immaterial is what your ICP and opportunity in the market are based on the features you have. As we start to see real usage patterns coming from agents we're going to have to rethink the role of the product manager. Is it more important to pixel push the perfect 15 screens a human needs or to pump out functionality that opens up usage of the product to a much larger swath of agents (say SMBs all the way to large enterprises). While more features does not necessarily mean better product, it usually means expansive ICP. Will software vendors shift this way? Will we unleash AI coding agents and build massive feature sets into our MCP layers that would never exist as human UI? Or we will limit our MCP to the same curation of what we show to a human? Time will tell.