AI has become incredibly prevalent since it became more publicly available just a few short years ago. Its integration to everything from translation to image generation — and especially in the areas of technology and programming — has made it a central part of many people’s processes.
I won’t deny that tools like ChatGPT and Copilot have helped me with the work that I carry out for some of my clients. In particular, checking and revising complex code has become less of a burden, which has, in turn, allowed me to work more efficiently and waste a lot less time in trying to find and fix problems.
Contrary to some people’s experience, though, where AI tools are used to write all of the code for a project, I use the experience I’ve gained over the last thirty years building and maintaining websites and browser-based web applications to continually become more efficient. Thanks to the routines I establish and continually improve, my clients know that when they ask for projects to be built, this doesn’t necessarily mean a huge budget. In fact, the prices of my websites have actually come down over the past few years, thanks to improved tooling, more knowledge, and a continued passion to deliver the best possible result.
Improvements in WordPress – in particular the Block and Site Editors – have allowed me to begin working much more efficiently than I used to at previous employers, and they allow clients to maintain their websites much more easily. By going “all-in” on the Block Editor when it became available at the end of 2018, I can now count over seven years of experience in building websites with this marvellous solution. However, leaning on these content management tools haven’t obscured the main aim of my work: helping clients to communicate with their audiences efficiently and working in an editorial environment which has been customised precisely for their needs.
The blog posts on independent websites that I mentioned earlier have reminded me to mention the downsides which using AI can bring. Thanks to the amount of experience that I have in building websites, I only really need to use AI in order to fix problems which aren’t immediately evident when I look at complicated code. Even then, AI can quickly get lost in the weeds when trying to solve a problem, because it won’t admit when it doesn’t know the answer. Instead, it just goes round and round to try and give a user an answer which they’ll accept, instead of actually solving the problem. The acceptance of the answer is the goal, not the solution of the problem.
AI isn’t any better in this regard than it was two years ago. While this may change in future, recognising this problem and working around it is a key part of my daily working life.
Over the past year or so, I have found that using AI to try and solve a problem has occasionally led me to waste an inordinate amount of time. When I’ve written my own code and asked AI to write code for the same purpose as a comparison, I’ve found that not only is my own code usually better than the code produced by the AI, but it’s also more readable and understandable.
Herein lies is a key difference: maintenance becomes easier on an ongoing basis because I (or other developers) understand the code which has been written. This is particularly important because many of my clients have been coming to me for support on their projects for several years and so it’s critical to ensure that I can look at code from some time ago and immediately understand what’s happening.
Where so many solutions are designed to get a project up and running quickly, this is usually at the cost of long-term clarity and maintenance. In particular, this recent blog post by Nathan Wrigley has highlighted the fact that using automatic tools to complete tasks can in fact be detrimental to our lives — not just programmers but also people in other walks of life where repetitive work is part of the experience. In Nathan’s story, company employees changed their processes to be more efficient, but overlooked the ancillary and key aspects to their working lives.
They did not know this at the time, but with hindsight, their efficiency drive had taken away something from them, something they regretted letting go of. But it was now gone, and there no hope of a return to how it was.
Nathan Wrigley, “Dangers of efficiency with AI”
Think of the programmer who sits in a home office alone all day, with no personal interaction, and uses AI tools to build projects which become so complicated that they are no longer maintainable six months down the road. Compare that experience with a developer who sits in a room with a team of people working together, or one who learns a programming language to completely understand topics like technology, accessibility and usability.
Although the former is probably more efficient at the beginning of a new project, the mid to long-term life of the project won’t necessarily require less work. It’s just that the work will be trying to work out what your AI built six months ago and why it isn’t working as you expected it to. Given recent experience, the AI will possibly even struggle to understand what it wrote six months ago.
Ask an AI to write a piece of code, and it’ll quite often write the code differently every time it makes an attempt. In comparison, the code I wrote for a website for a local paragliding school in 2019 is still running with no problems, over six years later, thanks to a deep understanding of the requirements and a commitment to the best possible approach and standardised and repeatable working practices.
Having the knowledge to understand the media in which you’re working, and learning how to build a more efficient process, is much more sustainable, ecologically friendly, and better for your career and life than using somebody else’s automated tool to help you out in the short term. Code is only the beginning of a successful project, because human requirements and ideas and situations are continually changing.
Mass-produced and AI solutions often sound attractive thanks to their promise of making work easier, but if you’re using AI, don’t forget to take a step back and look at the bigger picture. Will the work stand up for six years? Will you even understand it a year from now without any AI tools to explain it to you?
