Don't blame the tools, adopt them

Ian Ritchie / The Herald Business HQ / May 2025

WE HAVE ALL heard that Generative Artificial Intelligence (AI) is going to destroy lots of jobs – how accountants, lawyers and other administrative workers would soon be replaced by intelligent computer systems. It’s claimed that these new AI systems will be more reliable and accurate than the humans they replace.

But I was a little bit surprised to discover that one of the top prime targets that is emerging for AI to take over human tasks is that of computer coding, begging the question: Will we need as many software engineers in future?

This was a shock as it was my original choice of career. As an 18-year-old school leaver I was considering what I might study at university. My best grades were in English and Modern Studies, but I decided to follow a more vocational course that would ensure that I would always be gainfully employed, and I elected to study the emerging field of computer science.

Back in 1968 only two universities in Scotland were offering degrees in Computer Science: St Andrews and Heriot-Watt. I looked at the syllabus of St Andrews and concluded that it was a thinly disguised Applied Maths course, so I went to Heriot-Watt where I discovered – that it was also a thinly disguised Applied Maths course.

Never mind, it set me off on a career which, as I had hoped, I would always have a decent job. When I graduated with my computer science degree in the mid-1970s my timing was spot-on – the personal computer was just emerging, and PCs went on to re-build the computer industry from the ground up.

The next 20 years or so were a great time to be a software developer, as we created the modern graphics based personal computer and connected them up to the networks that form today’s hugely powerful global information environment.

But these days for the software developer would now seem to be over.

Generative AI has attracted a lot of coverage for its ability to ‘create’ creative material, whether its writing fiction, new music, or TV scripts. But in practice, AI systems are best suited to writing computer code.

Although creative AI usually falls down on actual creativity, coding doesn’t require creativity – it requires precision and accuracy, which are exactly the characteristics which are ideal for AI.

A whole raft of products have been developed to service this market. The currently most popular AI coding assistant, GitHub Copilot, is a collaboration between GitHub, OpenAI, and Microsoft. GitHub estimates that 92 per cent of US-based developers now use AI coding systems.

Other established coding tools are available from Anthropic, Meta, and Google; which are now vying with a vast number of start-ups most of which are raising huge amounts of venture capital.

ReflectionAI has raised $130 million, Anysphere has raised $105 million, and there is Poolside, which has raised a staggering $500 million at a $3 billion valuation.

Last month, the big fish in this pond, OpenAI, released a suite of new products, including GPT-4.1, 03 and 04 which are targeted at software development, and announced an AI agent called Codex CLI specifically designed to help users with coding tasks.

OpenAI’s chief product officer said: ”this is the year that AI becomes better than humans at competitive code”

Developments in this field are fast moving, in 2023, studies from an industry test called SWE-bench found that AI systems could solve 4.4 per cent of software problems, but they estimate that this has now risen to 69.1 per cent. A recent report from consultants McKinsey shows that software developers can complete coding tasks up to twice as fast with generative AI.

And it is already having an effect on employment. In January this year, Mark Zuckerberg announced that he was cutting back on software developers saying he planned to replace midlevel engineers at his company with AI systems.

Zuckerberg has concluded that AI can just take over some key coding tasks currently done by human hands.

Around the same time, Marc Beniof of Salesforce declared that they were not hiring any new software engineers for the whole of 2025. He claims Salesforce has seen a 30% productivity increase in engineering due to the use of AI technologies.

There are currently about 90,000 software engineers employed in Scotland, and the trade body ScotlandIS forecasts that that number will grow by over 23,000 each year, many more than our higher education sector are generating.

This scarcity means the average salary of software engineers is currently over £50,000 per annum.

But ScotlandIS’s recent annual technology survey reports that Scottish companies are currently taking a cautious approach to using AI.

It must be time to throw caution to the wind. The solution to today’s shortage of skilled software developers is will certainly lie in adopting the latest AI development tools.