It's no secret that generative AI (GenAI) has taken the world by storm since its launch in late 2022. Broadly, GenAI is perfectly suited to assume tasks that are monotonous, repetitive, or easily automated, making it an effective tool for improving employee satisfaction and reducing burnout. These capabilities are especially promising when you consider Gartner's estimate that up to 30% of the U.S.'s working hours could be automated.
Of course, GenAI can do much more than simple automation. It has brought out the best of human ingenuity, both in its original development and in the ever-expanding list of use cases ranging from creative art to intense data analysis. According to LTIMindtree's research, early adopters of GenAI have strategically implemented the technology to improve customer experience (81%), optimize processes (68%), and enhance product innovation (57%).
Similarly, DevOps revolutionized the software development and IT operations landscape since its inception in the late 2000s. Today, it is widely adopted across industries, transforming how organizations build, deploy, and manage applications. Its principles of automation, collaboration, and continuous improvement have become essential for staying competitive in the fast-paced digital world. Co-pilots that assist across various software delivery life cycles (SDLCs) like code generators, synthetic data generators, and more have already become engineers' best friends. However, challenges related to hallucination and accuracy still hinder the full realization of transformational benefits in DevOps. Knowledge graphs, small LLMs (SLMs), and AI agents will prove to be especially pivotal for technology leaders. DevOps implemented using these AI native techniques has the potential to deliver the promise of resilient and efficient IT along with speed to innovate.
Knowledge graphs
Knowledge graphs structure data into interconnected entities and relationships, enhancing AI's contextual understanding, data integration, and explainability. Roughly in the past 20 years, approximately 2.8 trillion lines of code have been written, which is over five times the estimated number of stars in the Milky Way. Each line of this code has both structured and unstructured data that is linked with it - user stories, defects, test cases, documentation, tickets, or diagrams. With the advent of GenAI, creation of these knowledge graphs is much easier.
By mapping out relationships between different pieces of data, knowledge graphs help AI systems gain a deeper contextual understanding, improving the accuracy and relevance of their outputs. Knowledge graphs make AI systems more interpretable by clearly showing how conclusions are derived from the data, which is essential for building trust and ensuring compliance with ethical standards.
Fit-for-purpose small LLMs
Any enterprise has its own nuances on processes, data that it harnesses within its specific industry as well as how various applications are built. For example, in an oil and gas enterprise, there will be specific workflows on how IT incidents are handled. Similarly in the financial services industry, the written code needs to comply with specific standards. The solutions built on top of out-of-the box LLMs do not have this context and suffer from hallucinations and inaccurate responses.
This is where if enterprises can harness the power of knowledge graphs and bring that contextualization via RAGs; accuracy improves multifold. This technique, also called GraphRAG, uses a knowledge graph as context for an LLM to generate text. It allows for more structured and contextually rich information to be incorporated into the generated text. Enterprises can create fit-for-purpose, contextual small LLMs for better solutions.
AI agents
An AI agent is an autonomous program designed to perform specific tasks. These agents use advanced algorithms and machine learning techniques to interact with their environment, make decisions, and learn from experiences. They are built to manage complex workflows, streamline processes, and make decisions autonomously. For example, in a customer service environment, agentic AI can handle inquiries, monitor customer satisfaction, and adapt responses based on real-time information. In a DevOps context, enterprise can have special agents that serve a specific purpose.
In typical operations, incident triaging takes significant time and needs human interventions. A triaging agent that draws insights from the small LLMs and takes action based on the incident situation can significantly improve the resiliency in an operations scenario. Similarly, a product owner agent can help recommend if an application needs to be enhanced or sunset based on the different DevOps and IT ecosystem parameters. Legacy modernization agents leveraging the Graph RAG-based SLMs can help in modernization of applications at an accuracy up to 70%-80%, which is significantly better than out-of-the-box genAI solutions can achieve.
GenAI is just getting started
The integration of GenAI with AI-native DevOps is revolutionizing the software development landscape. GenAI's ability to automate monotonous tasks and enhance creativity has significantly improved employee satisfaction and productivity. Early adopters have leveraged GenAI to enhance customer experiences, optimize processes, and drive product innovation. Similarly, DevOps has transformed IT operations with its principles of automation, collaboration, and continuous improvement.
Despite these advancements, challenges like hallucination and accuracy issues still exist. However, the combination of knowledge graphs, small LLMs (SLMs), and AI agents offers promising solutions. Knowledge graphs enhance AI's contextual understanding and explainability, while small LLMs provide tailored, context-specific insights. AI agents autonomously manage complex workflows and decision-making processes, improving operational efficiency and resilience.
By embracing these AI-native techniques, organizations can achieve more resilient, efficient IT operations and accelerate innovation, ultimately realizing the full potential of GenAI and DevOps in the SDLC.
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]