I’ve been dreaming about the future of software development for a while now. And I’ll admit, my vision might sound like a nightmare to some of you. But hear me out.
Picture this: You write a single prompt. Something beautifully simple like “Make me a text editor like Microsoft Office Word, just make it better.” You press enter, grab a coffee, maybe have lunch with your team, do some team building exercises.
When you come back, there it is. A perfect product waiting for you.
The software anticipates everything. It understands how humans interact with technology. It guides users through complex tasks seamlessly. All that’s left is calling the customer, showing them your creation, and sending the invoice.
That’s my dream. And yes, I know we’re nowhere close to that reality yet.
The harsh reality of generative AI in software development
Working with generative AI tools today feels vastly different from that utopian vision. Whether you’re using them for software development, marketing, or any other domain, you’ve probably noticed the gap between promise and reality.
The limiting factors are significant. Let me walk you through what’s holding us back.
Implicit knowledge remains our biggest challenge
One of the most significant limitations of generative AI isn’t technical—it’s human. Much of what makes great software comes from experience and intuition, not documentation.
In practice, this creates a gap:
- Developers rely on implicit knowledge built over years
- Teams understand users and business context intuitively
- This knowledge can’t be fully documented or standardized
Unwilling stakeholders create real barriers
Then there’s the human element. You might have stakeholders in your organization or your customer’s organization who simply say no. “Don’t use AI. We want you to write everything by hand.”
These concerns often come from valid places. Security worries keep people up at night. What if that beautiful new text editor you deployed leaks all user data to an unsecured S3 bucket somewhere on the web?
These aren’t hypothetical concerns. They’re real risks we need to address.
Business complexity overwhelms current AI capabilities
Modern businesses operate with complex data structures that often confuse even employees who’ve been there for years. How can we expect AI to navigate this complexity when humans struggle with it?
And then there’s the most critical challenge: Dependencies. Software components need to work together seamlessly.
When you tweak component A, you might break component B. Zoom out to the enterprise level, and you see external systems, APIs, processes, and compliance requirements all interconnected in ways that current generative AI simply can’t handle in one go.
Andrew Lovell
For expert advice like this straight to your inbox every other Friday, sign up for Pro+ membership.
You’ll also get access to 300+ hours of exclusive video content, a complimentary Summit ticket, and so much more.
So, what are you waiting for?
