Anant Deshmukh, CTO and Head of IT at ICICI Prudential Asset Management Company, on AI readiness, the 2026 CIO agenda, agentic AI, autonomous agents, and more.
AI is no longer an experiment. It has become a core enabler of business outcomes across industries. Yet implementing AI at scale, particularly in regulated sectors such as financial services, remains complex. As CIOs look toward 2026, they must strike the right balance between technology adoption, data architecture, operational scalability, governance, and regulatory compliance, while continuing to drive innovation.
Jatinder Singh, Editor of CIO & Leader, spoke with Anant Deshmukh, CTO and Head of IT at ICICI Prudential Asset Management Company, about how enterprises are navigating this shift. The conversation explored not only ICICI Prudential’s AI journey, but also the broader technology priorities and focus areas shaping CIO agendas for 2026.
ICICI Prudential Asset Management Company is a leading player in India’s asset management industry, operating in a highly regulated and scale-intensive environment. At the organization, Anant Deshmukh is responsible for shaping and executing the technology agenda across digital transformation, cloud adoption, data platforms, and DevOps. His role spans oversight of core and mission-critical applications, as well as building technology teams focused on scalability, resilience, and operational reliability.
Edited excerpts from the interview follow.
CIO&Leader: Implementing AI at scale is one of the biggest challenges for enterprises today. From your perspective, what architectural and technological hurdles must enterprises overcome to ensure AI is both effective and compliant?
Anant Deshmukh: Scaling AI across an enterprise is never just about technology—it’s about making everything work together seamlessly. Sure, data security, encryption, and access control are table stakes, but the real headaches come from privacy, regulatory compliance, and integrating AI with legacy systems.
Take Data Subject Requests, or DSRs, for instance. A customer might ask to erase or update their personal data, but that same data could already be part of analytics workflows or portfolio management systems. Architecturally, you can’t just delete it, you need a way to move it to compliant storage without breaking downstream analytics. Getting this right takes careful planning, investment in data lineage, strong governance, and smart workflow orchestration across the enterprise.
CIO&Leader: That seems like a significant operational and financial commitment.
Anant Deshmukh: Absolutely. Beyond technical infrastructure, enterprises must focus on auditing, vendor management, and policy enforcement. Every data change must be tracked and proof of compliance maintained. With AI, this also includes responsible AI governance, models must be ethical, explainable, and auditable.
The CIO agenda for 2026 is about scaling AI responsibly, integrating it end-to-end, and ensuring governance without stifling innovation. CIOs who master this balance will define their organization’s digital future.
CIO&Leader: How is ICICI Prudential approaching agentic AI and autonomous agents?
Anant Deshmukh: Agentic AI is evolving rapidly, but full autonomy will take time. Right now, we focus on orchestration platforms that manage multiple agents simultaneously, ensuring coordinated execution while humans remain in the loop.
Lessons from RPA are instructive: early automation projects often failed because bots lacked context and couldn’t handle dynamic workflows. Agentic platforms are different, they need multi-agent orchestration, exception management, and governance built-in.
From an enterprise perspective, deploying agentic AI isn’t just about buying software. It requires integration with legacy IT, embedding controls for compliance, monitoring agent performance, and training teams to manage them. The objective is end-to-end enterprise workflows that are largely automated, yet responsible and auditable.
Agentic AI is evolving rapidly, but full autonomy will take time. Right now, we focus on orchestration platforms that manage multiple agents simultaneously, ensuring coordinated execution while humans remain in the loop.
CIO&Leader:From your experience, what aspects of data architecture most directly impact the success or failure of AI initiatives in regulated environments?
Anant Deshmukh: Data is the starting point for everything we do with AI. We have built a lakehouse platform using a medallion approach, with bronze, silver, and gold layers that bring together transactional, demographic, and behavioral data into a single foundation. This structure allows teams to work with raw data, refined datasets, and analytics-ready information without losing context. We use internal GPU infrastructure to run machine learning models for advanced analytics, and we also enrich our models with external datasets to deepen insights.
However, the real challenge goes beyond storing data at scale. What matters just as much is data integrity, ease of access, and meeting regulatory requirements. If the underlying data is inconsistent, poorly governed, or difficult to trace, AI models will struggle to produce insights that can actually be trusted or acted upon. This is why CIOs need to take a holistic view of data readiness, focusing on control, quality, and traceability as much as on platforms and tools.
CIO&Leader: How important is middle- and back-office modernization in supporting AI initiatives beyond customer-facing use cases?
Anant Deshmukh: Critical. AI adoption fails if back-office systems can’t handle high volumes. Systems for onboarding, settlements, transaction monitoring, and fraud detection need to be scalable. Investment must be end-to-end, not just in analytics or customer-facing tools.
CIO&Leader: How do you balance centralized governance and decentralized execution of AI initiatives?
Anant Deshmukh: Our approach is centralized evaluation with decentralized execution. The central team defines standards, evaluates models, ensures compliance, and maintains oversight. Functional business units own specific AI use cases, allowing innovation closer to the business context.
This prevents bottlenecks while maintaining regulatory and operational discipline. It also encourages new ideas, as teams feel empowered to experiment while remaining under central oversight.
CIO&Leader: When you step back and look at the 2026 horizon, what investment priorities should CIOs be prepared to defend at the board and budget level?
Anant Deshmukh: CIOs need to invest in:
Data transformation: Consolidating and cleaning transactional, behavioral, and demographic data.
AI infrastructure: GPUs, orchestration platforms, and agentic AI frameworks.
Operational systems: Middle- and back-office scalability, trade and transaction monitoring.
Governance and compliance: Auditing, responsible AI frameworks, regulatory alignment.
Talent and skills: Embedding ML engineers, AI architects, and data scientists across business functions.
Responsible and ethical AI has to be non-negotiable. Models must be explainable, auditable, and aligned with regulatory expectations. This is essential not just for compliance, but for building long-term trust with customers, regulators, and internal stakeholders.
CIO&Leader: Finally, when shaping the AI agenda for 2026, what are the most important considerations IT leaders must address to move from pilots to enterprise-wide value?
Anant Deshmukh: IT leaders need to approach AI as a true enterprise-wide initiative, not a collection of isolated experiments.
First, responsible and ethical AI has to be non-negotiable. Models must be explainable, auditable, and aligned with regulatory expectations. This is essential not just for compliance, but for building long-term trust with customers, regulators, and internal stakeholders.
Second, organizations must be operationally ready end to end. AI cannot sit on top of weak back-office systems. Core functions such as onboarding, settlements, trade monitoring, and fraud detection all need to scale in step with AI-driven decision-making.
Third, governance needs the right balance. A centralized team should define standards, guardrails, and compliance requirements, while individual business units drive execution. This model allows innovation to move quickly without losing control.
Data readiness is another critical foundation. CIOs must ensure data is consolidated, clean, and easily accessible. Investments in lakehouse platforms, medallion architecture, and scalable infrastructure are no longer optional if AI is to deliver real value.
Finally, talent remains a decisive factor. Embedding data scientists, machine learning engineers, and AI architects across business functions enables faster experimentation, smoother prototyping, and more effective adoption across the enterprise.
