The Idea is Valuable Again
There is a sentence that I, as an innovation manager in a corporation, have said a hundred times: “The idea alone is worth nothing.”
This sentence has an expiration date. And it’s approaching faster than most people think.
Why Ideas Were Worth Nothing for a Long Time
The principle was simple. Ideas are everywhere. What mattered was the ability to execute them. And this ability depended on things that most people didn’t have or couldn’t control.
I’ve seen this play out in three variants over and over again.
The wrong environment. A colleague in an IT corporation had a great idea for an IoT alarm button in offices. Elegant solution, real need. But we were a B2B IT service provider. The product didn’t fit the portfolio, the leverage was missing, the idea disappeared into a drawer. Not because it was bad, but because the place was wrong.
The missing skills. A hairdresser noticed that all booking systems for his salon were either incredibly expensive or incredibly bad. He had a clear vision of how it could be better, even a specific feature set in mind. But he’s a hairdresser, not a software developer. The idea is good. Yet, it remains without consequence.
Bad timing. You’re early, you have an idea before the market is ready, and two years later, you see someone else becoming successful with it. I know this from my own experience as well.
The pattern behind all three cases: Ideas always needed an execution machine behind them. And not everyone had that machine at their disposal.
What’s Shifting Right Now
I was skeptical for a long time about the speed at which AI agents are changing software development. I have since revised this skepticism.
My current assessment: Within this year, 50 to 70 percent of code will be written fully autonomously by AI agents. Not because new technological breakthroughs are necessary, but because existing scaling paths are being followed consistently and the tools that already exist today are being introduced step-by-step. One more model generation that delivers today’s performance at lower costs, and that point is reached.
What this means: The execution machine is suddenly available to everyone.
Our hairdresser describes his idea for a better booking system to an AI agent. The agent develops the product. End-to-end. Without developers, without VC funding, without a corporate sponsor.
The Idea as Complete Input
It gets even more interesting when you think this through further. AI agents can do more than just take over the programming process. They can also fill roles where humans previously lacked the skills: product management, operations, support, marketing & sales.
In product management, this means: continuously capturing user feedback, prioritizing feature backlogs, adjusting roadmaps. In sales: formulating initial outreach, optimizing onboarding copy, evaluating conversion metrics. In marketing: producing content, conducting SEO analyses, managing campaigns.
What’s emerging is a world where a precise idea is actually enough to build and market a product. Because execution is no longer tied to human capacity, skills, or institutional power.
Who Profits Today
This world is not yet accessible to everyone. The first wave is software engineers. They know the infrastructure, understand the tools, know where automation runs smoothly and where they still need to step in manually. They are the first who can operate as one-person product teams.
The first commercial products show where this is leading. Tools like glazeapp(dot)com promise to build desktop apps in minutes through a conversation with AI. Not yet a complete product team in a box, but the direction is clear: the idea as complete input, the agent as the execution layer, the product as output.
In the months and years to follow, others will join. The hairdresser. The craftsman. The freelancer. Anyone who knows a real problem and is willing to learn the tools.
Where This Leads
Before we get too euphoric, it’s worth looking at something that sounded almost the same a few years ago: the democratization of content creation.
Suddenly, everyone could write professional texts, generate images, produce videos. The barrier to entry fell away. And what happened? The volume of content exploded. Attention did not. The distribution of reach followed a power-law curve: a small number of creators reaped the majority, while the broad mass in the long tail remained practically invisible. More producers, same “winner-takes-most” dynamic as before.
There is little reason to assume it will be different for software products.
Our hairdresser builds his booking portal. And 500 other hairdressers do too. All with similar ideas, all with the same agent toolset. Most of these products will hardly be used because they lack a clear differentiator, a compelling distribution strategy, a community, or a genuine market moment. Built by someone who had the idea but didn’t understand deeply enough why customers buy or switch.
So, what’s changing: the execution hurdle is falling. What isn’t changing: products that truly dominate markets come from people who deeply understand a real problem, address the right users, and provide a concrete reason why their product specifically should be chosen. That is product management in the truest sense. And that was never a question of programming.
The question is therefore not: Can everyone build a product now? The question is: Who builds products that are actually used?
What Remains
The sentence “The idea alone is worth nothing” was never wrong. It was a description of reality under certain conditions—namely, that execution was expensive, slow, and tied to scarce resources.
These conditions are changing fundamentally right now. The execution gap that separated ideas from products for so long is closing. What becomes more expensive at the same time is user understanding. Market intuition. The ability to name the actual problem behind a problem before starting to build.
Anyone who has a precise idea today, based on a real problem and addressing the right users, has more than ever before.
Have you already tried to implement an idea with AI agents? And where did you notice that the actual work only begins when the building stops? Feel free to write it in the comments.