Explore the delirious rantings of Frederick Vanbrabant. A blog focused on the intersection of Enterprise Architecture, product, and business strategy.

Frederick's delirious rantings

AI won't automatically speed up software development processes because the real bottleneck isn't coding speed — it's the quality and completeness of inputs that developers receive. Drawing on lessons from 'The Toyota Way' and 'The Goal', the argument is that AI-assisted development still requires detailed, well-defined problem specifications from domain experts, which is exactly what developers have always needed. True process improvement means ensuring bottlenecks receive predictable, high-quality inputs rather than simply adding more resources or tools to the bottleneck itself.

I don't think AI will make your processes go faster

I agree with the point author is making, and here’s an example from my experience. Below is a part of the software development lifecycle (SDLC) that focuses on quality and the shift-left testing approach (this is the actual SDLC followed by one of the companies I worked for):
<ol>
<li>Requirements are collected.</li>
<li>Designs are created.</li>
<li>Designs are reviewed and approved.</li>
<li>Business specifications created.</li>
<li>Business specifications reviewed and approved.</li>
<li>Technical specifications are created.</li>
<li>Technical specifications are reviewed and approved.</li>
<li>Test cases are written.</li>
<li>Test cases reviewed and approved.</li>
<li>Implementation is written.</li>
</ol>
If you have an SDLC like this you’re already ready for AI writing the code for you because steps 1-9 will result in detailed specs where everything adheres to standards and was validated multiple times at different levels. But how much of an AI productivity boost will you get? Not much, because you still have steps 1-9 that actually consume most of the time.
If you have a SDLC like this, you’re already prepared for AI to write code for you, because steps 1-9 will result in detailed specifications that meet standards and have been verified at various levels. But how much of a productivity boost will you get? Not much, because steps 1-9 still take up the majority of your time.
Without AI, if you needed to ship faster (for example, because of a tight deadline), you’d cut cornets on steps 1-9 (or simply skip some of them), and the implementation would still be fine because the developer would have context (it will be in his head and not in the specs) and could figure things out on the fly. With AI we need to complete steps 1-9 because otherwise there will be not enough context which can result in poor implementation (the same thing can happen when code is written by a human, but AI is more vulnerable to it).
All this means that whether you use AI or not, if you want to improve productivity, you need to address the SDLC as a whole. And if you use AI, you must ensure that every stage of the SDLC is optimized for it

<blockquote>
“Every software developer knows that you can’t make projects go faster just by typing faster. If that were the case we would all be taking typing lessons.”
</blockquote>
Well said! I hope more people would recognize this especially now that the public’s sentiment on AI is currently nosediving.

Everyone think about AI as a replacement for someone or some process (step of development). It doesn’t work that way.
But if you think about at AI as helper at every stage, performing small tasks and removing some of the routine, it greatly speeds up the process.
So yes - AI **_will make your processes go faster _**but u need use it in right way.
hey google please build module to send sms - wrong
hey google please create method to normalize phone number - correct