- DevOps is an early test case for the potential of AI to separate tech leaders and laggards
- Top‑performing DevOps teams automate functions like testing and spend more time building new features
- Automating these tasks allow developers to act more like business consultants
To understand how artificial intelligence will widen the gap between top‑performing IT teams and the rest of the field, take a look at DevOps, the software development methodology that many IT leaders now swear by.
While DevOps teams have used automation tools in software testing for years, recent advances in machine learning and AI are yielding significant gains in productivity and efficiency.
The highest‑performing DevOps organizations, measured by factors like deployment frequency and change failure rates, have fully embraced AI tools and automated 72% of their development processes, according to a 2017 Puppet survey of over 3,200 DevOps professionals.
These top performers do significantly less manual work than low performers. The survey found that compared to laggards, DevOps leaders have automated 33% more of their configuration management, 27% more of their testing, 30% more of their deployments; and 27% more change approval processes.
By integrating AI into DevOps, these forward‑thinking companies aren’t just accelerating production cycles and catching product bugs and security flaws earlier. They’re empowering developers to expand their skills and assume valuable new roles.
“Automation enhanced with machine learning has completely changed the industry,” says Sylvain Kalache, co‑founder of Holberton School, which trains full‑stack software engineers.
When DevOps met AI>When DevOps automation met AI
Since it first emerged a decade or so ago, DevOps has offered a new approach to software development, one based on intense collaboration between developer and operations teams in order to build, test and release software at far faster rates than was possible in previous years.
Tech bellwethers like Amazon, Google, and Netflix were early DevOps disciples. Now thousands of other companies, including American Airlines, Hertz, and Nordstrom, have adopted the practice. In a 2017 Forrester survey of software developers, 90% reported that their companies were using DevOps in some form, or have plans to start.
The benefits of automation to DevOps processes have been obvious from the beginning, says Michael Azoff, principal analyst at research firm Ovum. By eliminating manual steps, organizations can reduce human error in app delivery while speeding up the development process.
Emerging AI tools stand to generate even bigger gains, says Stephen Elliot, VP of management software and DevOps at International Data Corp. Specifically, AI will be key to automating two critical processes: continuous integration (CI) and continuous deployment (CD).
In a CI process, code is integrated into a shared repository several times a day with the goal of detecting problems. In a CD process, code changes are automatically built, tested and prepared for release to production.
By using AI algorithms to automate both of these vital processes, DevOps teams can test for potential bugs and security flaws early. This speeds up product releases and frees developers to focus on more strategic tasks.
At USA Today, for example, developer teams release 110 targeted editions of the publication each day for more than a dozen mobile devices. This required 588 separate UX tests managed by DevOps teams, according to TechTarget.
Until last year, tests took 90 minutes to complete. But after USA Today implemented an AI‑enabled testing platform for CI and CD, the average test time shrank to less than ten minutes.
Similarly, when HP’s firmware division invested in automated testing for its LaserJet 4, the DevOps teams were able to redeploy staff and boost time spent developing new features sevenfold, according to the Puppet report.
Such efforts “are driving operational effectiveness and better ways of working,” Elliot says. They also speed time‑to‑market cycles and boost product quality.
Schooled by AI
Nascent AI tools for software development are also changing how schools teach young developers. At Holberton School in San Francisco, students train to become full‑stack software engineers through hands‑on, team‑based learning. AI tools are already an integral part of the curriculum.
One reason for the shift: Most of Holberton’s teachers and mentors are working industry professionals who are acutely aware of the significance of AI‑powered automation for the future of DevOps.
“The idea behind these tools is to get machines to ‘self‑heal’ without human intervention,” says Kalache, the school’s co‑founder. “The more data you have, the more you can predict the problem that your infrastructure, service or server might have and how to automatically fix it. AI and machine learning are at the heart of what makes these tools efficient.”
One consequence of this new approach is that many developers are taking on more strategic roles in their organizations.
“Developers are becoming more like business consultants—they sit next to their product’s ends users and figure out what they really want from the product,” says Torsten Volk, managing research director of IT consulting firm Enterprise Management Associates. “A good developer is one who understands the customer. The best developers are those who listen the best.”
A growing trend that Volk points to is software vendors bringing developers into their support centers to engage with customers.
What are customers’ frustrations? What would make the app work better for their needs? Developers can use these insights to refine existing features and dream up new ones.
Kalache finds that some IT roles are becoming obsolete. “Roles like system administrator are becoming a thing of the past,” he says. That’s because there is no longer a need for tasks such as capacity planning and performing manual and repetitive work. “The new breed of system administrators [are] automating themselves out of the job, so that they can focus on innovation rather than operations.”
Developers who dabble in AI needn’t worry about job security if they can upgrade their skills. “We used to hire developers with specialized skills,” Volk says. “Now we hire full‑stack developers who do everything from database to business processes to the presentation layer—they understand the whole process. They understand the impact of the software before they even start coding.”
The more AI frees DevOps teams from mundane tasks, the more it empowers them to help companies outpace the competition in speed, security and innovation.