Monday August 4th 2025 by SocraticDev
First part of a series on integrating artificial intelligence into DevOps practices
I believe the formalization of the DevOps engineer role (or DevOps specialist) really matured alongside the rise of cloud computing.
One of the key value propositions from public cloud providers was streamlining the traditional system administrator (sysadmin) role, along with related positions like network technicians. Commercial cloud platforms present themselves as pre-built resource blocks that are optimized out-of-the-box and often managed by the cloud provider under a defined SLA. The idea was that you'd just need to train developers or retrain willing system administrators to bridge the gap between programmers (dev) and operations (ops).
Generative artificial intelligence - AI that can generate text, code, and images - is disrupting this whole concept. The programmer's supremacy is being challenged; chunks of code can now be generated in seconds. Same goes for operations: mysterious problems can be debugged and resolved in a fraction of the time thanks to ChatGPT - it's proven.
Why are DevOps folks best positioned to become "AI natives"?
While DevOps isn't really a profession so much as a culture around tech products, I believe today's DevOps practitioners will be the first to become AI natives. Why?
First, because DevOps people generally have an opportunistic mindset. They love solving problems and are naturally inclined to set aside preconceptions to adopt simple, effective solutions like process automation and "everything-as-code."
Second, being a DevOps engineer is less about a job title and more about a state of mind. By pivoting toward AI, DevOps practitioners reinforce this mindset, which boils down to first-class pragmatism.
Most importantly, the main objective of DevOps is the software development lifecycle (SDLC). It's about taking ideas and transforming them into products and services that add value for customers. Everything else - infrastructure-as-code, automated pipelines, collaboration with developers - these aren't ends in themselves.
The primary goal is always to deliver quality products or services to people who are willing to pay to use them. And that won't change with AI.
How to pivot toward AI: a pragmatic approach
It's funny, but the idea of embracing artificial intelligence as a DevOps practitioner is still controversial. We love to mock AI suggestions when the real culprit isn't the tool but the person using it.
For me, pivoting toward AI doesn't mean blindly delegating all our tasks to ChatGPT. It means first accepting this tool by learning how to use it better.
- Master the art of prompt engineering: Understanding how to optimize our questions to get better answers.
- Develop critical judgment: Knowing when to use ChatGPT and when to take time to think and carefully read the documentation.
It's a bit like when you need to accomplish a task that takes 2 minutes once every 6 months: is a good DevOps engineer going to spend a week automating that task, or pass on it?
Technical foundations: where to start?
To learn how to better use AI, you need to start with the fundamentals.
Since we like to deliver value quickly, I suggest skipping a few chapters of AI history and jumping straight into today's hot topic:
Natural Language Processing (NLP) and Large Language Models (LLMs). These two concepts represent the foundation of revolutionary generative AI tools. Understanding how ChatGPT works will help us better exploit it while knowing how to manage its shortcomings.
Recommended learning resources
There's a ton of online resources for getting familiar with NLP and LLMs. From magazine articles to university courses, I'd like to choose free options that include hands-on work and some form of certification.
Hugging Face's Natural Language Processing Course (free) with exams
Conclusion
I've proposed the idea that DevOps practitioners (or DevOps engineers) have professional traits that favor pivoting toward a world where artificial intelligence has a primary role in designing and delivering tech products and services.
As a call to action, I'm proposing to follow two introductory tracks offered by reputable AI organizations: HuggingFace and Google. Understanding the technology behind revolutionary tools like ChatGPT will immediately allow us to use them better and then continue our pivot by digging deeper.
It's still early, but I believe we'll then study the libraries and frameworks used by industry players (e.g., TensorFlow, PyTorch, and scikit-learn), then explore data management aspects: data wrangling, ETL, and preprocessing with tools like Pandas, NumPy, and SQL. Then why not demonstrate our knowledge by diving into workflow automation for real use cases involving models and external APIs?
To be continued...
translated from french by Claude Sonnet 4
