To begin talking about agentic artificial intelligence (AI), the first step is to establish what it means and how it’s different from Generative AI (gen AI). Other than its fantastic content generation capabilities, gen AI can help orchestrate workflows involving multiple tools and components, within the bounds of human-defined rules.
Agentic AI takes this a step further by introducing agents that can handle complex, multistep tasks in real time by autonomously reasoning and deciding the best course of action. The concept behind agentic AI is to emulate “digital workers” which can learn, reason and act with minimal human prompts.
Reach out to discuss this topic in depth.
Fig 1: Agentic AI workflow
The next question that needs to be addressed is how mature is agentic AI currently and is there any real uptake for it in enterprise scenarios? As is the case with any emergent technology, it’s important to distinguish between reality and hype.
While the current landscape can be described as emerging, it’s also maturing at a fast pace. In terms of enterprise deployment, agentic AI is mostly at a pilot or early deployment stage with productivity benefits expected to accrue over 3-year of 5-year deal terms.
Providers currently are developing proofs-of-concept which they can showcase to Enterprise buyers. While we have seen some co-investment taking place in these pilot projects, in many cases, providers are absorbing the cost of such pilots- in the hope that they can scale these solutions later.
Use cases for agentic AI in cloud and Infrastructure services
So, are there any real use cases for agentic AI in cloud and infrastructure services as of today?
In some recent deals, we have seen agentic AI take an increasingly prominent role in Service provider solutions.
Analyzing the ticketed and non-ticketed effort across L1, L2, L3 operations makes it abundantly clear that in a modernized estate (virtualized, templated, patched regularly), agentic AI can add significant productivity gains on top of the gains accrued from traditional automation and gen AI (see Fig 2 for some commonly articulated use cases).
Unlike traditional script based/ runbook automation and gen AI led solutions, which add value to individual points of the overall workflow, agentic-AI can fundamentally reimagine complete workflows (through the systems of execution approach, where multiple AI agents work in harmony to execute a task end-to-end.)
The coup de grâce of agentic AI is autonomous reasoning and decision making, which means that not only does it automates tasks, but it also decides on how to do it, through a self-correcting mechanism.
We already see evidence of service providers articulating huge productivity figures over a 3-year or 5-year deal term (by virtue of AI agents replacing a very significant chunk of human Full Time Equivalents (FTEs)). Today, more than ever before, it is imperative for stakeholders to have a very clear picture of use cases, productivity potential and Return on Investment (ROI) associated with these solutions- if they want to get ahead of the curve and unlock full value from contracts.
Fig 2: Impact of Agentic AI on Compute/Server Operations
Limitations to a fully autonomous AI agent future- human-in-the-loop design
If agentic AI is so powerful, will it not replace all human FTEs at some point in the future? Concerns over data security, reliability of AI decisions, and regulatory compliance are leading to hesitance among Enterprises around adopting agentic AI wholesale.
There is emphasis laid on human-in-the-loop designs for the proper governance of agentic AI models. This ensures that the final sign-off and accountability of agentic AI workflows rests with a human agent. So, while, in theory, agentic AI technology allows for full autonomy, in actual practice, a human-in-the-loop design ensures predictability and accountability of outcomes.
The second half of 2025 will see agentic AI become more mainstream. Success for Enterprises will depend on selecting the right service provider partner who can help unlock value from AI-augmented service contracts.
If you found this blog interesting, check out our Agentic AI In CXM: Early Signals On What It Is And What It Is Not | Blog – Everest Group, which delves deeper into another topic regarding agentic AI.
If you have any questions or want to discuss the evolution and commercial impact of agentic AI on cloud and infrastructure services in more depth, please contact Saikat Roy ([email protected]) and Ricky Sundrani ([email protected]).
