When should you NOT use AI in your workflows?
When speaking to friends and peers, the topic of AI comes up more often than perhaps any of us really want it to. We all know that the impact over the last few years has been drastic and difficult to ignore. The overarching rhetoric I normally hear is, I would say, overwhelmingly negative. In my circles at the very least, people are concerned about the impact on the environment and human creativity. It’s hard as someone entrenched in the technology industry to disagree with this, but it’s also impossible to ignore the gains that it can bring to myself and my team.
The tech world is competitive. 10 years ago, running a software business in the UK and fielding an entirely UK based team put us at somewhat of a disadvantage. It meant that for our service to be competitive we had to provide value above and beyond what could be provided in other areas of the world at a cheaper price. I am proud to say that Vivato managed to do that and excelled throughout the years, keenly aware of our position in the market and the forces that governed that position.
Things have changed since then and the requirement for entry into the tech world has plummeted which in my opinion is not a bad thing. The vibe coders have risen up and the world of code is more accessible than it has ever been. The ever present LLMs are adept at explaining key concepts that might have taken a frustrating hour of googling a few years back, even then meaning you had to sift through hundreds of stack overflow threads (those that have done so know that posting on stackoverflow itself is rarely advised).
So when these AI tools presented themselves, the Vivato team jumped on the bandwagon and suddenly the amount of time spent copying and pasting to create tedious data files was cut by a huge margin. It was exciting and we have kept up with the lightning fast changes and features that have been produced by the major players. There still is though the conundrum of when and where it is right to use.
This question has multiple layers. The reality is that AI and the ferocious arms race going on all over the world for the fastest and most impressive models IS having an impact on the planet, with data centres being erected seemingly overnight and sucking up all natural resources around them (including our precious RAM cards). There’s not really an argument to be made that the impact of this technology has been good on the planet.
It is easy to argue that a piece of AI artwork is soulless and has most likely been trained on data stolen from creators that deserve credit that corporations won’t be giving them. People want humans to be involved in the creative purpose and for there to be some meaning and passion behind what they are consuming. There is a lot less desire in the world though for artisan software, I appreciate good software architecture as much as the next person in my industry, but unfortunately they don’t tend to fill galleries with database diagrams. What this means is that including AI in our workflows isn’t really a choice. You either do it or you will be left behind.
So what’s our current strategy? AI in moderation, and with the core tenet being human understanding and decision making. AI tools can help take away a lot of monotonous work but there needs to be oversight and understanding behind the implementation. We don’t include any artificial workflows in our code and deployment review processes, even though this is being pushed hard by a lot of the big players. We consult AI tools when designing architecture but our tech leadership team is in charge of the decision making.
It’s a balance that is constantly evolving and changing as new features are being released each week. I have some members of my team that I have to keep from testing each shiny new tool that comes out and some team members that occasionally need to be (gently) reminded that they don’t need to do everything by hand anymore.
I do miss the days of not having to question the veracity of everything I see online, having one less existential dread to push down, and this specific topic dominating every other conversation. It is nice to not have to manually input so much data though.
Let me know your thoughts on the matter and whether it’s possible to take an ethical objection to AI as a software developer today
No AI was used in the creation of this blog post!
