Coding Me, Coding You: Reflections from the Golden Window
Focusing on more than just AI, it addresses software production, quality expectations, and our transforming functions, as code generation becomes less of a constraint.
Downstream processes are now the bottleneck.
This comment by Joost Noppen from BT Group really struck a chord.
When AI generates code faster than we can clarify needs, integrate adjustments, or validate performance, the primary hurdle becomes the surrounding ecosystem, not the code creation itself.
This is a rallying cry for me, given my passion for Quality Engineering and strengthened a conviction I’ve had for some time:
- Quality is not a hurdle; it is a system and a culture.
- In an AI-accelerated world, this system needs to be dynamic, adaptive, and always learning.
- Requirements shift, as do client expectations.
Clarity and precision are now more crucial than ever before in the AI world. Passive test strategies and once‑a‑year quality plans just won’t cut it.
- We require living artifacts that adapt alongside the project.
- We require more rapid feedback loops.
- Address ambiguity at its origin, not after it becomes embedded within the product.
Another big takeaway from Joost’s talk was that BT generates around 12 million lines of AI-assisted code monthly, yet they implement only about 30%.
Joost made an interesting argument: he envisions a future where AI agents are as essential to engineering as Excel is to business today.
I can subscribe to this point of view.
The commoditization of code generation brings it closer to manufacturing, once again bringing into sharp focus my experiences in FMCG and mass production.
- Think process control.
- Think monitoring the “production line.”
- Think continuous improvement.
It’s amusing how lessons learned during my manufacturing career influence my perspective on AI-driven software development today. I never imagined at the time that I would find it so useful in this context!
The legal discussions led by Debora Dorn from Appleyard Lees IP LLP were equally fascinating. It made me draw parallels with music copyright. While the number of notes is limited, the quantity of songs is vast. Similarity differs from infringement.
- Context matters.
- Interpretation matters.
- Intent matters.
It served as a reminder of the intricacies and nuances of generative AI and the ongoing new for legal framework evolution.
The event left me feeling energised. Not because AI will solve everything. But that it drives us to pose better questions regarding:
- What defines quality?
- How can we optimise human contribution and influence?
- How do we reshape working practices to complement this unfamiliar landscape?
All before the “golden window” of opportunity closes.
If anything, I’m more persuaded than ever that Quality Engineering will remain relevant.
Software development and quality practices are changing, and those who engage with it early will define what follows. I, for one, aspire to be part of that.