In this interaction, Venkatesh Thenkarai, Chief Delivery Officer, Bahwan CyberTek, discusses how enterprises can harness agentic AI to transform IT operations and decision-making, build AI-ready data ecosystems, optimise cloud investments through FinOps and intelligent operations, and balance innovation with sovereignty, security and regulatory requirements.
Agentic AI is rapidly emerging as the next evolution of enterprise automation. How do you see autonomous AI agents transforming cloud management, IT operations, and enterprise decision-making over the next few years?
I think the important shift with agentic AI is not that machines will become more conversational. It is that enterprise systems will increasingly move from detecting problems to deciding what should happen next and, within defined boundaries, executing the response themselves. That changes IT operations considerably.
Today, a large part of cloud and IT management still follows a chain of monitoring, alerting, ticket creation, human diagnosis and human intervention. Agents can compress that loop. They can correlate events across infrastructure, applications and business systems, identify probable causes, recommend an action and eventually execute approved remediation.
We are already putting this model into practice through BCT’s AiGenix managed services partnering with iFIX Tech. The combination brings agentic intelligence into service desk, application management, infrastructure, end-user computing, platform engineering and database services. It is designed to move operations from reactive support towards proactive, self-healing and outcome-based service delivery.
What makes this interesting is that it is not about replacing the ITSM environment. The platform orchestrates across existing ITSM, ERP, CRM and enterprise systems, with autonomous triage, contextual decision intelligence and self-healing execution, while escalating situations that require human judgment.
We already have 5 plus live enterprise programs across ITSM and shared-services transformation, with activity across North America, India, the Middle East and APAC.
And I think the next phase is particularly interesting, where we intend to take this beyond horizontal IT operations into industry-specific agentic solutions for BFSI, oil and gas and retail, while also co-developing multi-agent platforms that can coordinate across business functions.
That is where I think agentic AI becomes much more consequential – not just an agent that performs a task, but a system of agents that can understand context, make decisions and coordinate execution across an enterprise.
But autonomy without controls is not enterprise technology. Identity, permissions, policy boundaries, observability, audit trails and human escalation have to be designed into the architecture.
Data continues to be the foundation of every successful AI initiative. What are the biggest challenges enterprises face in building trusted, integrated, and well-governed data ecosystems that can truly power AI at scale?
The biggest data problem enterprises have is not a shortage of data. It is a shortage of clean data they can trust enough to act on.
Most large enterprises have accumulated data across applications, business units, geographies and generations of technology. The problems then become predictable – different definitions for the same metric, duplicated records, inconsistent master data, inaccessible legacy systems, unclear ownership, limited lineage and data silos.
These problems get more pronounced with AI initiatives with the underlying data which, is fragmented or poorly governed, AI makes decisions faster with unreliable inputs.
Our case experience makes this very clear. In one banking engagement, restructuring the MDM environment reduced duplicate customer records by 8% and invalid data by nearly 30%. In an oil and gas environment, standardising master data across rigs and integrating Maximo and Oracle created a unified view and reduced dependence on manual intervention.
I see the data foundation as five layers – data quality, context, governance, access and integration. And increasingly, I would add a sixth i.e. AI readiness. Enterprises need governed pipelines, common business definitions and real-time access so that AI can work with business context rather than simply raw information. The companies with most data may not necessarily get their AI implemented right but those, with the clearest line from data to decision.
As cloud investments continue to grow, organisations are under increasing pressure to optimise costs while maintaining agility. How are practices such as FinOps, workload optimisation, and intelligent cloud operations helping enterprises strike this balance?
I would like to first clarify that cloud optimisation is not about finding cheaper infrastructure.The right question is: are we engineering the workload correctly for the business requirement?
A workload that runs 24/7 when it only needs peak capacity intermittently is an architecture problem. Paying for oversized compute is a workload-design problem. Duplicated tooling and unnecessary data movement are integration problems. FinOps brings visibility to the bill, but engineering has to act on that visibility. That is why I see FinOps, architecture and operations becoming much more tightly connected.
The next step is intelligent operations. Once you bring together cost, performance, utilisation, architecture and operational data, AI can start identifying patterns that humans would otherwise have to find manually. The direction we are taking with our intelligent operations capabilities is to move from static dashboards that tell you what happened to systems to predict future budget run-rates, actively discover idle resources and execute real time downsizing during low traffic periods.
I don’t see FinOps as “how do we spend less on cloud?” I see it as “how do we make every unit of cloud spend correspond to a business requirement?”
With increasing concerns around data sovereignty, cybersecurity, and regulatory compliance, how should enterprises rethink their cloud and digital infrastructure strategies to build resilience while maintaining control over critical workloads?
I don’t think enterprises should frame this as a choice between cloud adoption and control over critical workloads. Critical workloads need a deliberate architecture around data residency, encryption, identity, access, auditability, regulatory requirements and disaster recovery. Some workloads may belong in a public cloud. Others may require private infrastructure, sovereign environments or a hybrid architecture.
The mistake is treating the cloud decision as a migration exercise rather than an architectural one. We have seen this in regulated environments. In healthcare, for example, BCT built a cloud-native platform while working within clinical compliance requirements such as HL7 and FHIR.
We have also worked on environments where resilience had to be designed into the architecture, including disaster recovery with an RTO of one hour and an RPO as low as five minutes. And as we move towards agentic operations, this becomes even more important. If an AI system is going to make and execute operational decisions, enterprises need to know what that agent is allowed to do, what data it can access, why it made a decision and how that decision can be traced.
So sovereignty is not about keeping everything inside one physical boundary. It is about knowing exactly where critical data and workloads are, who can access them, how they are governed and how the enterprise continues operating when something goes wrong.
Bahwan CyberTek has built strong capabilities across industries such as banking, energy, government, and supply chain. From your perspective, how important is deep domain expertise in ensuring that AI and cloud transformation initiatives deliver meaningful business outcomes rather than becoming purely technology-led projects?
Deep domain expertise matters because enterprises don’t buy technology in isolation. They buy a change in how a part of their business works.
That distinction shows up across BCT’s work. Our case portfolio spans across BFSI, government, oil and gas, energy, lifesciences, retail, logistics and Tele Com industries across geographies. The technology patterns may be similar, but the consequences are not.
In banking, an AI system may influence credit or risk decisions. In energy, it may influence asset availability. In government, it may affect access to essential citizen services. In supply chain, it can directly affect working capital and service levels.
That is why the combination of technology capability and domain expertise is so important. For me, domain expertise changes the question from “Where can we apply AI?” to “Which decision in this business is worth making better, faster or earlier?” That is where technology starts becoming a differentiator.
BCT has introduced platforms like CloudXcel and agentic managed services to enable more autonomous technology operations. How do you envision these intelligent platforms changing the way enterprises manage cloud environments, improve operational efficiency, and accelerate innovation?
I see these platforms helping us move from managing infrastructure to engineering conditions under which infrastructure can manage more of itself. Cloud environments have become extremely sophisticated, but that sophistication has also created fragmentation.
The engineer is still expected to connect those silos and decide what to do next. That is the thinking behind the evolution of BCT’s agentic managed services. Through F3AI (Fit For Future) framework and the iFixGYAANi platform, we are bringing autonomous intelligence into multiple managed-service towers rather than treating AI as another tool sitting on top of operations. The platform can perform autonomous triage, contextual decision-making and self-healing execution, while integrating with the enterprise systems already in place.
And we have evidence that this is moving beyond a concept. In one large managed-services deployment, we have reduced monthly incidents by 23%, turnaround time by 30%, and improved first-call resolution from 70% to 86%. In another North American retail deployment, ticket resolution improved by 50% and agent workload fell by 30%.
The future direction is even broader. We are looking at industry-specific agents and multi-agent systems that can coordinate across functions rather than operate within a single IT workflow.
The important word for me is control. Autonomous does not mean unsupervised. A mature platform should know what it is allowed to change, why it is making a change, what evidence supports that decision and when it should hand control back to a human. This is probably the most practical path to AI-native operations: don’t replace the enterprise; make the enterprise more intelligent.
Large-scale public-sector digital transformation projects often become benchmarks for enterprise modernisation. From BCT’s experience of working on complex government and public-sector platforms, what key lessons can organisations learn about building future-ready digital infrastructure, modernising data platforms, and executing transformation at scale?
Large-scale transformation is ultimately an exercise in getting technology, data and operations to work as one system.
At scale, the complexity is rarely just technical. You are dealing with multiple systems, large and constantly changing data volumes, different stakeholders, stringent governance requirements and infrastructure that has to remain available while it is being transformed.
One of our government engagements illustrates this well. BCT implemented an AI-powered observability framework for a high-traffic state e-Service platform handling more than two million daily transactions. The platform achieved 99.9% uptime across 50,000-plus concurrent users, while response times improved by 60%, application errors fell by 75% and MTTR reduced by 45%.
The lesson is that future-ready infrastructure cannot simply be about adding new technology. It has to make the entire environment more observable, predictable and responsive.
The second lesson is to build the data foundation properly. We have seen across large enterprises that inconsistent data, disconnected systems and legacy integration can become the biggest constraint on modernization. In one oil and gas engagement, for example, we created standardised master data and connected Maximo with Oracle to establish a unified view across operations.
And finally, transformation at scale has to be designed for continuous evolution. You cannot build a platform today assuming that the business, data volumes or technology landscape will remain unchanged.
For me, the measure of successful transformation is therefore not the number of systems modernised. It is whether the new architecture gives the organisation better information, greater resilience and the ability to respond faster to what comes next.