The real opportunity for enterprises lies in Agentic AI

This is a guest blogpost by Arunava Bag, CTO, Digitate

Enterprise adoption of Artificial Intelligence (AI) is no longer a question open to debate. Last year, PwC’s June 2025 Value in Motion report found AI could boost global GDP by up to 15% by 2035.

The right question to ask is, if, according to MIT, 95% of generative AI pilots at companies are failing, how should enterprises adopt AI successfully at scale? A 2024 Digitate report also revealed that while 92% of organisations had implemented AI and automation, they had underdeveloped or limited strategies in place to maximise that investment. It seems companies are betting (and betting big) on AI, yet nearly all enterprise pilots are stuck at the starting line.

For IT operations in particular, AI has a lot to offer, ranging from detecting anomalies, suppressing alert noise, automate the last mile to resolve the issue and predict future incidents. Still, the question of how organisations can fully harness this potential, and with which form of AI, is only beginning to be explored.

The case for and against GenAI

Generative AI has opened up a wide range of possibilities. Various industry verticals are harnessing the power of GenAI for creative content generation, efficiency improvement, and personalisation of experience. IT operations are no different either. Generative AI offers significant potential in enhancing operational knowledge processes, including pseudo-code development, automated documentation, test-case design, and test-data generation, among other advanced functions. According to a 2024 Forbes Advisor poll of 165 respondents, the vast majority (79%) have used generative AI (such as ChatGPT) to help them at work. However, while co-pilot style AI may help proof or check work, it doesn’t change how an organisation actually operates – the workflows and manual steps remain there. Complex tasks remain human touch intensive. Only Generative AI will only take you so far.

This is where Agentic AI comes in, as a new evolution in enterprise systems. One in which technology no longer simply reacts to alerts, but can perceive, analyse, and act independently, guided by continuous learning from real-world outcomes. Simply put, given a complex task, an agentic AI system can understand the context, orchestrate reasoning steps, take action to close the loop and learn in the process, improving the next iteration, without much human intervention for most of the time. Agentic AI operates independently (within defined boundaries), making informed decisions without requiring constant human supervision or step-by-step instructions. It fuses intelligence with automation in a way that redefines how workflows, and pushes IT into a new era of proactive, adaptive, and autonomous operations.

The agentic manifesto

Agentic AI extends the boundaries of intelligent automation by autonomously understanding context, making decisions, taking action, and learning continuously. These systems work alongside human experts to navigate unfamiliar scenarios, adapting and generalising from experience. With the advancement of Generative AI and specialized AI agents, unknown tasks encountered for the first time also becomes possible up to some extent, with decent degree of confidence These AI capabilities make it ideal for addressing some of the biggest IT challenges that enterprises face, while driving greater agility and efficiency.

For instance, an AI Agent for IT incident Management employs an agentic framework to establish a closed-loop operational model. Simply put, this system autonomously perceives the intent and context, reasons the possible solutions, acts to close the last mile, and learns in the process for an enhanced process next time around. It does so by integrating diverse agents, stores/access relevant data, and analytical tools into a coordinated, end-to-end process. All of this, while orchestrating them into a unified flow.

Unlike traditional systems, AI Agents can autonomously handle most routine IT events in the background, while also delivering conversational intelligence at the front end to enable meaningful interactions with human experts. Each agent has a specialised role, collaborating like members of a well-coordinated team. With AI Agents, SRE and ITOps teams can shift from reactive firefighting to proactive issue detection and continuous improvement.

The question of ‘how’?

AI adoption is not meant to be a siloed process within a specific department or unit; rather, it should be a holistic approach. Siloed adoption is inconsistent, and can mean different departments advance faster than others, when what enterprises are looking for is overall transformation. This is one of the major advantages of Agentic over Generative AI. In some cases, there are long learning curves. Siloed deployments repeats this long learning curves, rather that benefitting from the perspective already gained.

There is also the ‘ongoing challenge of measuring AI’s impact on productivity and profit’. How can enterprises accurately measure the effectiveness of AI when it’s implemented on an individual basis? If AI is only deployed in silos, how can enterprises truly gauge its effectiveness? Without broader visibility into how it’s being applied in practice, there’s no clear way to assess its productivity or overall value.

With AI being adopted in isolated pockets, organisations eventually reach a ceiling. There are only so many use cases that can be executed within a single department, which naturally limits growth. In addition, a meaningful scale soon requires data, inputs, and workflow integration from other business functions. This is something siloed adoption simply can’t support. It will only take you so far.

For a powerful technology like AI and especially agentic AI, guardrails are extremely important. Though early days, frameworks (e.g. TRISM) have become prominent. As the AI/agentic AI systems are so powerful and trusted, security also becomes a very crucial point. Organisations should not lose sight of these aspects while undertaking enterprise-wide adoption.

Through the adoption of autonomous operations, enterprises are advancing toward a future in which operational intelligence is both automated and autonomous. Where systems don’t just notify teams about issues, but actively prevent them, enabling humans to focus on innovation instead of firefighting. The MIT report reveals that while most revenue-growth AI initiatives are currently failing, this doesn’t mean the opportunity is any less important. How can companies capitalise on this? Leverage Agentic AI to transform the way we work.