Turning creators into curators: AI implementation in DevOps
Generative AI has the potential to boost productivity in key parts of the technology sector, largely by automating repetitive and time-consuming tasks, with McKinsey projecting it could add between $2.6 trillion to $4.4 trillion in economic value each year.
Software development is one particular area where automation and AI could make a huge difference. Due to the current digital skills shortage and the long, arduous DevOps process, pipeline automation could – if done right – save businesses time and money.
However, implementing AI is generally not straightforward, and if not done carefully, it risks backlash from developers and customers alike. Amazon showed how this can go wrong, with recent AI-led reforms causing backlash from engineers.
Stress increased, standards and morale dropped; and this was just an internal issue. When new tech rollouts impact customers or draw attention from regulators, the financial and reputational impact can be worse.
President and CEO of CloudBees.
Both the potential gains and risks are increased for large enterprises. As the vast majority have at least some of their tech stack on-premises, and most innovation is targeted at cloud-based Software-as-a-Service products, enterprise pipelines often get left behind. To make matters worse, they are also harder to modernize due to the scale and complexity of the organizations’ pipelines, and there is more to lose, with far-reaching reputational and regulatory risks.
To avoid this upheaval while capitalizing on productivity gains, enterprises should avoid broad implementations. Focused, lower-risk, and clearly defined problem areas are key – particularly the automation of code testing and issue prioritization, which is both the main source of developer friction and the simplest to automate.
Addressing developer toil
Developer productivity and morale is the most valuable resource for enterprise IT teams, and also the one facing the most strain. AI has the potential to be a game-changer for addressing this ‘developer toil’; providing developers with more time to focus on creative tasks rather than mundane and repetitive ones.
Arduous and repetitive tasks and high workloads, do more than demoralize. Developer toil causes delayed projects, poor performance, and unsustainable staffing levels as developers quit – further contributing to an industry struggling to find and retain talent. In 2024, developer toil was reported as the reason for team members quitting by over half (52%) of developers.
In software development, the main culprit for creating this toil, and therefore the priority for automation, is in the ‘post-commit to production process’, also known as ticket creation. AI can be used to automate the triage process – quality assurance (QA), continuous integration (CI), and vulnerability management – by categorizing, grouping, and prioritizing failures without human assistance. This frees up valuable time, and makes sure this time is spent on the most pressing issues in the software pipeline.
Priorities in AI adoption
Most enterprises are very open to using AI, with almost half of technology leaders in a 2024 PwC survey reporting that AI was “fully integrated” into their companies’ core business strategy. It can be done, but it must be done correctly, and having technological eyes bigger than your operational stomach can lead to data privacy and governance concerns, alienated staff and customers, and ultimately a slowed digital transformation.
Scope is essential. Targeted applications, focused on protected ‘sandboxes’ without access to direct outcomes minimizes risk and allows the process to be better observed, learnt from, optimized, and then rolled out further. Businesses must also keep in mind that outsourcing a whole generation of software to AI code generation risks more than just poor outcomes; any ‘black box’ will make it impossible to diagnose and fix potential errors down the line.
The future of AI in DevOps
In the future, AI has the potential to transform pipelines into intelligent, self-optimizing systems with better powers of prediction and iteration. For now though, as across all other sectors, it must go hand-in-hand with close human oversight.
The most important part of any AI adoption is still, and forever will be, the humans working alongside and overseeing it. As with all tech rollouts, staff need sufficient training and the ability to feedback any issues with team structure, or the technology itself, to uphold morale and to make best use of the new solution itself.
Furthermore, any issues around developer toil won’t be fixed if AI is only used as an excuse to work DevOps teams harder.
We list the best sites for hiring developers.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro