Kaius Meskanen is a founder and entrepreneur with a passion for innovation, leadership and the future of work.
Building a functional mobile app traditionally costs between $30,000 and $700,000, with most projects taking three to six months to complete. Additionally, many major apps are developed over a time span of years.
Those numbers are shrinking fast, yet most businesses haven’t adjusted their modernization strategy to match that pace. I’m not talking about AI making developers slightly more productive. I’m talking about nontechnical founders building production-ready applications in days using tools like Cursor, Replit Agent and Lovable.
The existential question in software has officially shifted from “Can we build this?” to “Should we build this?”
What Actually Changed
Three things happened in the last 18 months to make conversational app development viable.
First, LLMs got really good at generating working code. GPT-4 and Claude can now create entire application structures with debugging, security considerations and scalable database design. They’re trained on massive codebases and know patterns that used to require tenured developers. According to McKinsey research, generative AI accelerates certain coding tasks by 35% to 45%.
Second, no-code platforms improved beyond simple form builders. They can now quickly put together complex applications from pre-tested modular components. AI acts as the one who makes decisions by choosing and configuring these pieces based on natural language descriptions.
Third, cloud services have made important components accessible. Features such as security authentication, payments, data storage and push notifications now have standards, allowing AI to integrate them automatically instead of needing bespoke development.
What we get as a result is the following: Describe what you want in plain language, and AI converts it directly into functional code, database structures and user interfaces.
The Implication Nobody’s Talking About
Speed comes with a caveat. Sure, it helps save money, but it also completely changes what’s strategically possible.
Consider a company testing a new feature. The traditional process involves assembling a team, planning sprints, building for months and finally releasing to users. When each test requires an investment, you can only afford a few attempts before someone demands ROI.
GitHub’s research on Copilot found that developers completed tasks 55% faster with AI assistance. But the real advantage beyond individual productivity is the ability to run multiple experiments simultaneously.
When testing costs drop dramatically, you can run 10 experiments for the cost of one. A retail company wondering about a mobile loyalty program can build a working version in days, release it to a test group and make decisions based on actual usage rather than surveys and focus groups.
What This Means For Startups
The traditional startup funding model exists largely because building technology is expensive. You need significant capital to create something users can interact with, which means raising money before you have proof that your idea works.
That equation is flipping. Founders now reach product-market fit with minimal capital, then rise from a position of strength with real user data. The value of a technical co-founder has shifted from “someone who can build” to “someone who knows when to build versus when to test with AI-generated prototypes.”
The Skills That Really Matter Now
If AI handles implementation, then what becomes valuable?
- Problem Identification: The hardest part of building successful products has always been figuring out problems that need solving and for whom, rather than coding or development. When you can build anything quickly, choosing what to build becomes the real differentiator.
- Strong Product Sense: This involves understanding which features drive the most engagement, which offerings users truly need and how to plan a product roadmap that accounts for these learnings. AI is nowhere near being able to make these calls yet.
- User Experience Design: AI generates functional interfaces, but making something delightful still requires a human eye and gut. The next best skills will be the ability to identify valuable problems and opportunities and to create experiences that genuinely serve user needs.
Real Limitations To Keep In Mind
AI-generated code gets the job finished, but it won’t win any architecture awards. If you’re building for millions of users or handling serious data complexity, have a human developer review what the AI produces. For internal tools and most business apps serving a smaller group of people, the AI output may work.
Here’s AI’s kryptonite: edge cases. AI builds for the scenario where everything goes right. Inconsistencies—like a user entering weird data, for example—expose bugs that need human attention.
AI trained on common frameworks and popular patterns does well with standard applications. Try building something truly ground-breaking, and you’ll likely need developers to handle what the AI doesn’t know.
Therefore, start with lower-stakes projects. Build internal tools to pressure test AI applications. Create standalone apps before trying to integrate deeply with critical systems. Learn what AI handles well, then gradually expand the scope.
The Clock Is Running
The gap between AI capability and human-level development shrinks monthly. Capabilities that seem just beyond reach today are likely to be routine in six months.
But here’s what matters more: Organizations embracing these tools early are accumulating compounding advantages. Every application you build teaches you what works. Every user interaction generates data about what features matter. Every iteration refines your understanding of your customers.
Companies cycling through this loop rapidly are building knowledge and capability advantages that slower-moving competitors can’t easily match. The winners won’t necessarily have the most resources or the best technology, but they will be the ones moving from hypothesis to validated learning the fastest.
What You Should Do This Week
If you haven’t built something with AI in the last 30 days, you’re already behind. Not because the technology is perfect (it isn’t), but because your competitors are learning and you’re not.
Pick a real problem in your business—something small and annoying. Use Cursor, Replit Agent or Lovable and spend two hours turning an idea into a working prototype.
The point isn’t to create something useful, but to help you evaluate what’s possible.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
