The influence of AI in software engineering has had far reaching consequences for the enterprise, and it stands as a compelling example of a profession being revolutionized by new technology.
Sure, it’s nothing new to the sector, but the enthusiasm around generative AI in particular is speeding up the integration of AI across a wide variety of use cases.
AI looks set to impact all aspects of the software development lifecycle, from planning and designing, through to building, testing and deployment. We asked the experts how artificial intelligence is likely to shape software development, and what the biggest challenges might be.
1. Massive increases to speed and productivity
It's widely believed that AI-powered software will help developers and engineers across all stages of the lifecycle, which will have a huge impact on teams’ productivity. Many tasks will be automated, freeing up staff to work on more human-centric jobs and speed up processes.
It's believed that AI-powered software, named TuringBots by Forrester analysts, will be able to assist teams with – amongst other things – designing, coding and testing, taking the repetitive ‘grunt’ work off their hands.
Tools of this kind are already in play, such as Github’s Copilot and Amazon Codewhisper, and early signs suggest coders are enjoying a daily productivity rise. According Forrester vice president and Principal Analyst Diego Lo Giudice, this rise is as much as 20-40% in some cases, dependent on their skillset and the complexity of the program.
He predicts that this percentage will be the overall productivity gain across all areas of software engineering within the next two to four years.
2. Software engineers will need more diverse skillsets
The integration of more AI-powered software, or 'TuringBots', will likely lead to some roles disappearing, or at least evolving into something new.
Nicola Martin, head of Quality Engineering at Adarga and chair of the Software Testing Specialist Group at BCS, The Chartered Institute for IT, frequently advises those in the sector that they’ll need to diversify in response to roles being merged and team make-ups changing – something she’s seen first-hand.
“In my area, quality, it’s been a massive shock and in terms of the software testing industry things have been turned on their head. There’s been job losses and changes to the structure of teams,” she says.
3. Many software roles will be amalgamated
However, positivity is growing as software engineers are discovering new opportunities will enable them to branch out into different areas.
This is down to a growth in demand for AI-empowered applications across different departments, and AI model development shifting out from data science teams to software engineering, enterprise application teams and the rest of the business.
“We’re being encouraged to learn about large language models and do things like prompt engineering, and we’ve seen roles as diverse and strange as data scientist quality engineer,” Martin says. “There’s about four different skillsets wrapped up in that role: data science, machine learning, software engineering and testing.”
She believes opportunities for software engineers will arise in different teams as AI permeates a business, with roles in areas as varied as cyber security or marketing as departments use more AI-empowered applications.
4. Better software collaboration will be needed
As AI-powered tools become more accessible to software engineers, there will be a need for better coordination of how assets are developed and managed. This will lead to much closer collaboration between software engineers and AI developers – as well as the wider business – as departments begin using more AI-powered software in their day-to-day work.
According to analysts at Gartner, almost 80% of organisations have a formal/structured operating way of collaborating between AI teams and software engineering, and those that do are more likely to achieve success, sustain AI efforts across a number of business units and act strategically.
5. Diversity will improve to counter AI bias
Preventing AI bias will become much more important in the software engineering world due to evolving roles and growth in collaboration across teams.
Definitive guide to practical AIOps
Get guidance on how enterprises should be using AIOps today
DOWNLOAD FOR FREE
We’ve already seen many examples of the impact of AI bias, which is caused by models trained on datasets filled with existing biases against different sexes, races and other minority groups.
Red flags have been raised around the use of AI in recruiting, for example. Back in 2018, Amazon had to can an AI tool for being sexist, while in 2023 Workday was hit with claims that its AI hiring systems were discriminatory.
Then there’s the US government migration app, CBP One. This had issues recognising faces with darker skin tones, which stopped many people from being able to use the app to apply for asylum. The reason given was that datasets used to train the model hadn’t been diverse enough, with little data included on people with darker skin.
Diversity within teams that develop AI systems will be required to ensure that the final software doesn’t perpetuate existing biases, as diverse teams will help ensure AI models are trained on a wide variety of datasets.
“If you don’t have diversity at the beginning, you won’t get it in the end product,” Martin states.
6. TuringBot adoption rates will vary sector by sector
Not all industries will embrace AI-powered software at the same pace, and so the pace of change within the software engineering industry will vary sector by sector. One of the most common reasons for this is ongoing concerns around IP protection and security.
According to Forrester, many in the finance, manufacturing and pharmaceutical industries won’t use TuringBots that upload differentiating and competitive ‘crown-jewel traditional software’, with one global manufacturing CIO telling analysts that their business’ lawyers have prohibited the use of generative AI coding tools.
Improvements in secure and high-quality code generations will help increase trust and push adoption, but it will still take some time for certain sectors to embrace AI-powered software fully.
