Expert Insights

Alan Buxton proposes a fundamental rethinking of software perfection, insisting that AI-assisted software development doesn't need to be flawless to be functional and valuable.

He introduces the 80/90 rule as a viable and cost-effective alternative to complete solutions, arguing that it is often more important to quickly deliver good enough software that covers most use cases than to spend substantial time and resources perfecting every aspect of it.

Hear Alan explain:

  • The concept of the 80/90 rule, where 80% functionality can often be achieved at 90% less cost with the help of AI.
  • The real-world implications and benefits of faster delivery of “good enough” software.
  • The costly pitfalls of over-engineering software to cover all potential uses.
  • The importance of understanding the balance between error tolerance and accuracy, asserting that it is not a purely technical decision.
  • The shift to an AI-assisted development mindset that considers value and speed over traditional notions of software perfection.

Quote

quotation-marks icon
And the concept is with ai, mainly with ai, you can get 80% of the functionality of your swan key. A SaaS platform at 90% less cost. And so for a lot of use cases, that is actually quite good enough. The downside of the, the older way means that you end up with these... hopelessly over-engineered stuff. That cover covers every potential use case, which then becomes unusable so people don't use it. $400 and $4,000 are very different things. So you can't use something that thinks that they're similar to, to still solve those sorts of problems.quotation-marks icon
Alan Buxton ,
CTO, Simphony

THE NEW DEFAULT angle

There are clear opportunities to increase software development velocity and cost efficiency by embracing an 80/90 approach to AI-assisted workflows:

  • Evaluate your perfection targets: Identify where an 80% software solution would suffice and implement AI to achieve it quickly. Efficiency is often more important than covering every possible use case.
  • Understand the balance: Recognize the difference between need-for-speed and critical precision areas. Not all software components require absolute perfection. Use AI to deliver 'good enough' where it makes business sense.
  • Avoid over-engineering: A complex software solution that covers all potential use cases can often become unusable and unadoptable.
  • Address the psychological shift: Challenge the notion that less than 100% functionality equals failure. Communicate the benefits of early value realization and accelerated user feedback.
  • Embrace error tolerance: Opt for a fast, 80% accurate solution where feasible, rather than waiting for a time-consuming perfect one.
  • Iterate based on real user feedback: Use the early deployment of the software to gather meaningful user feedback for continuous improvements.
  • Effectively recognize and navigate the hype around AI: Understand that while AI can optimize many areas, it isn't a magical solution for everything. Apply realistic expectations to avoid disappointments or wasted resources.