With AI, what should we be focusing on?
Last week, Nicola, Lachlan, and I attended the Yow! Tech Leaders conference in Melbourne. The event was a delight, offering an opportunity to connect with old and new acquaintances.
Notably, several talks at the conference centered around the topic of Artificial Intelligence (AI). Given its widespread adoption across industries and within companies for both external product development and internal team productivity enhancement, it's understandable that AI was a prominent focus. Thus, we should be asking:
- How useful has AI been during the development phase?
- Are the teams more productive with AI?
- Can we achieve more with less?
- What is the effect on quality?
- What are the long term implications?
- How can we better use AI?
- How do we lead an AI transformation within our companies?
Key takeaways from the conference regarding AI are:
- Flow and productivity can be negatively impacted by substantial waste, regardless of code assistants (Sarah Taraporewalla)
- AI accelerates our existing behaviours and humans default to adding stuff when solving problems, because that feels like progress, and we don’t tend to subtract stuff. But constantly adding stuff creates complexity that we usually don’t need. Focusing on speed without ensuring quality creates systems that degrade over time (Annie Vella)
- Systems are prone to breaking when teams lack skills, intuition, or a clear focus on what makes up quality. The adoption of AI can exacerbate these issues (Annie Vella)
- Al-assisted codebases are getting bigger, faster. Added code increases; churn increases; and moved code goes to 0% (which indicates less refactoring). Good engineering practices still matter (if not more) and can mitigate Generative AI (GenAI) risks (Sarah Taraporewalla)
- Using AI tools results in a 4% boost in productivity, but only 1% of GenAI is committed without significant rework (Rasmus Lystrøm)
- Treat AI as a non-human system user by applying robust architecture, engineering patterns, validations, and guardrails to ensure effective communication and collaboration (Simon Wardan)
In essence, while AI holds the potential to boost productivity, the current benefits are not substantial enough. Moreover, it's essential to adhere to good engineering practices more stringently than ever before to mitigate GenAI risks. There are no free lunches when it comes to implementing AI into your work environment. Uncritical use of AI without good engineering practices results in unsustainable performance: your future self will hate you for the complicated spaghetti code your current (immature) AI is generating en masse. You can shortcut now, but it’s not a long term strategy!
How useful do you find AI at your work? Share your experiences with us.