While generative artificial intelligence (AI) has captured to popular imagination it’s becoming apparent that multiple forms of AI will soon be used in combination with one another to automate DevOps workflows at levels of unprecedented scale.
Machine learning algorithms, for example, are already employed by many organizations to automate everything from analyzing log data to identify security issues to surfacing patterns in tests made using multiple regression, function, and unit testing tools.
With the rise of generative AI, it is now also becoming possible to, for example, automatically generate code and create summaries of reports that can be more easily understood by multiple stakeholders. Generative AI doesn’t replace the need for previous use cases involving machine learning algorithms. Instead, generative AI makes use of a form of deep learning algorithms to create a different type of AI models. As a result, DevOps workflows will soon be invoking multimodal forms of AI to automate workflows in ways that should enable DevOps teams to automate processes at unprecedented levels of scale.
In fact, IT professionals that found DevOps too intimidating to adopt will soon find DevOps has become accessible enough to not just comprehend but also embrace. In effect, AI will drive the democratization of DevOps best practices as it becomes simpler to programmatically manage IT environments. At the same time, many DevOps professionals might event rediscover the joy of automation that attracted to DevOps in the first place.
Naturally, it may be a while before what is becoming a multimodal approach to AI becomes pervasively applied to DevOps workflows but now it’s more question of when rather than if it will become commonplace. The thing that each DevOps team will need to determine next is to what degree to rely on the AI models being developed by vendors versus either opting to customize an existing model using their own data or, more ambitiously, building their own.
There are obviously going to be tradeoffs. Organizations that opt to build their own models might see better outcomes but the level of data science expertise to build and maintain AI models is considerable. Others may opt to invest in a vector database that can be used to customize an existing AI model by exposing it to more current data. The issue that needs to be resolved is the degree to which organizations will want to build and maintain a vector database that is connected to a public AI model to achieve that goal when a managed service provided by a vendor provides the same capability.
The simplest path in most cases will be to rely on the managed DevOps service. One of the downstream impacts impact AI is likely to have its it will push more organizations to consume managed DevOps services that have been augmented using multiple classes of AI models. The cost of building and maintaining AI models requires a level of data science expertise that most DevOps teams are not going to be able to attain and retain.
There will, of course, be plenty of security and compliance issues that will inevitably arise. Not every organization is comfortable with the idea that the intellectual property that drives their business is hosted anywhere other than on IT infrastructure they control. Organizations that have those concerns are about to discover that in the age of AI that satisfying them is about to become a lot more expensive if expect to stay relevant as rivals take advantage of managed AI services to build and deploy higher quality applications faster.
In the meantime, DevOps teams would be well-advised to start determining what manual processes will soon be automated by AI today and tomorrow as the pace of innovation only continues to accelerate. AI isn’t going to replace the need for DevOps engineers any time soon but a lot of the toil that conspires to make DevOps more tedious than anyone cares to admit will soon thankfully for all concerned be eliminated.
