Sandeep Malhotra outlines why enterprise AI success depends on orchestration, governance, explainability, trusted data, and human oversight at scale.
Enterprise AI adoption is moving beyond experimentation into large-scale operational deployment across banking, insurance, healthcare, manufacturing, and enterprise services. However, as organizations scale AI, they are also confronting challenges around fragmented AI ecosystems, governance, explainability, legacy infrastructure, model management, and regulatory compliance. The industry conversation is increasingly shifting from isolated AI use cases toward building enterprise-wide AI operating models that are scalable, secure, and accountable.
In an interaction with CIO&Leader, Sandeep Malhotra, Chief Strategy, Solutions & AI Officer at Digitide Solutions. discusses why AI must be treated as a long-term organizational capability rather than a standalone deployment. He explains how Digitide’s Pulse.nerve orchestration platform addresses disconnected AI silos while emphasizing the importance of human oversight, ethical AI, explainability, domain expertise, and lifecycle-based AI governance.
CIO&Leader: Digitide recently introduced Pulse.nerve. Can you explain the orchestration platform and how it addresses disconnected AI silos across enterprises?
Sandeep Malhotra: Pulse.nerve is a subset of Pulse.AI and functions as a Model Context Protocol platform built around MCP servers across enterprise environments.
Traditionally, if one AI capability had to interact with multiple systems, organizations needed separate integrations for every interaction. As enterprises scaled AI, the number of integrations became unmanageable.
AI never operates in isolation. It works through data, platforms, integrations, enterprise ecosystems, and external capabilities. The real challenge is orchestrating all of them in a unified manner.
Pulse.nerve standardizes these interfaces, so AI capabilities only need to integrate once instead of repeatedly across different systems. Since AI never works in isolation and depends on data, platforms, integrations, and external ecosystems, orchestration becomes essential.
The platform essentially acts as the nervous system of our AI architecture. Standardized integration also improves governance, security management, and operational scalability.
CIO&Leader: Digitide works extensively with highly regulated sectors like BFSI. Where should organizations draw the line between AI autonomy and human intervention?
Sandeep Malhotra: In regulated industries such as banking and financial services, we deploy many AI capabilities on-premises using physical GPUs inside secure enterprise environments.
AI performs exceptionally well where organizations need speed, scale, data processing, pattern recognition, and cognitive correlation. However, human judgment still remains irreplaceable.
Human judgment remains irreplaceable. AI can analyse signals at scale and speed, but humans still have to decide the intervention, the accountability, and the outcome.
One example is our HR organization, where we use multiple AI agents across the employee lifecycle. One AI agent, Nikki, continuously analyses employee engagement signals, operational concerns, workload patterns, and workforce sentiment across nearly 55 ,000 employees globally.
The HR leadership team then uses those insights to make personalized decisions around job rotations, workload balancing, or employee interventions.
AI handles large-scale analysis, while humans remain responsible for judgment and intervention.
CIO&Leader: Many enterprises struggle with AI readiness and scaling. What is the most important question organizations should ask while assessing AI readiness?
Sandeep Malhotra: The most important question is whether AI is creating measurable enterprise value.
We do not approach AI as an isolated capability. We view it as a full lifecycle that begins with consulting and organizational readiness assessment. The future of work will involve humans and machines operating together, so enterprises must evaluate technology readiness, talent readiness, operational maturity, and governance simultaneously.
AI readiness is not a one-time exercise because AI evolves continuously. Organizations must constantly adapt models, workflows, and data strategies.
Responsible AI is also foundational. Ethical AI frameworks are critical because AI systems cannot be allowed to create unmanaged risks or unintended consequences.
Organizations create real value only when they treat AI as an end-to-end operational capability rather than a collection of disconnected pilots.
CIO&Leader: How does Digitide help enterprises manage model drift, operational costs, and legacy modernization challenges?
Sandeep Malhotra: Platforms like Anthropic, Devin, and GitHub Copilot are changing how enterprises modernize systems. Instead of rebuilding legacy infrastructure entirely, organizations can increasingly extend capabilities through AI agents and orchestration layers.
At the same time, AI systems are never static. Business requirements evolve, data changes constantly, and models require continuous retraining and optimization. Organizations also need safeguards against hallucinations, model drift, and model poisoning.
We therefore work with customers through long-term outcome-based engagements, where the focus remains on delivering measurable business value over multiple years while continuously managing AI evolution behind the scenes.
CIO&Leader: In BFSI, how much of AI-driven ROI comes from risk avoidance rather than direct operational savings?
Sandeep Malhotra: Both are equally important. Efficiency gains are often the easiest outcomes to achieve, especially in underwriting and decision-making environments.
For example, in crop or property insurance, AI can simultaneously analyze weather forecasts, soil conditions, geography, environmental data, and multiple risk variables. This enables insurers to price risks more accurately and improve profitability.
Similarly, in lending and collections, AI helps assess delinquency risks, determine customer behavior patterns, personalize repayment engagement, and identify the best communication channels for collections.
These systems improve operational efficiency while also helping financial institutions reduce non-performing assets and improve long-term risk management.
CIO&Leader: AI systems remain vulnerable to hallucinations and bias. How do those risks emerge in insurance and claims environments?
Sandeep Malhotra: Responsible AI becomes extremely important in insurance environments.
Problems usually emerge when organizations fail to build explainability, governance, unbiased decision-making, and trusted data sourcing into AI systems from the beginning.
For example, in parametric insurance modeling, organizations forecast risks such as floods, hurricanes, or wildfires. If important environmental variables are excluded from the model, the consequences can be severe.
During the California wildfires, several insurers underestimated how long fires could persist and failed to fully account for humidity, weather patterns, rainfall forecasts, and environmental conditions. That resulted in billions of dollars in losses.
This demonstrates why domain expertise is as important as technical expertise while building enterprise AI systems. AI engineers alone cannot solve these challenges. Industry specialists must also be deeply involved in the modeling process.
CIO&Leader: Many enterprises still operate on fragmented legacy systems. Can AI still deliver meaningful outcomes without complete modernization?
Sandeep Malhotra: AI fundamentally depends on access to data. Legacy systems themselves are not necessarily the problem. As long as organizations can access usable data, AI can still generate value even within fragmented infrastructures.
Legacy infrastructure is not the real limitation for AI adoption. The real limitation is whether organizations can access usable data from those environments.
We have worked with enterprises where data existed in completely different formats across systems accumulated over several years. Using AI-driven orchestration and normalization, we were still able to create unified insights and decision-making systems.
Even older manufacturing environments without modern SCADA capabilities can improve outcomes if data generation mechanisms are available.
The critical factor is not whether infrastructure is modern, but whether data is accessible.
CIO&Leader: BFSI is often considered a benchmark for governance maturity. What governance principles should other industries adopt?
Sandeep Malhotra: Ethical AI principles are now becoming table stakes. Organizations must prioritize vulnerability testing, security validation, explainability, trusted data sourcing, compliance management, and bias reduction from the very beginning.
AI can also significantly improve regulatory operations themselves. Regulations such as India’s DPDP framework require extensive privacy and consent management. AI can help enterprises intelligently manage customer consent, risk analysis, and compliance workflows at scale.
Many banking institutions still prefer running AI entirely within their own data centers and private GPU environments because governance and security remain top priorities. That approach may be more expensive, but it provides significantly stronger control and compliance assurance.
CIO&Leader: What is the biggest misconception enterprises still have about AI transformation?
Sandeep Malhotra: One of the biggest misconceptions is that AI is a standalone technology initiative. We evaluate AI through three lenses: industry, process, and personas.
Every industry has different AI requirements. Healthcare, BFSI, manufacturing, utilities, and retail all require different strategies and operational models.
Processes matter equally. Functions such as customer lifecycle management, order-to-cash, underwriting, and employee lifecycle management all present unique AI opportunities.
Finally, personas matter because the priorities of a CFO are very different from those of a CHRO or COO.
AI transformation succeeds only when organizations integrate all these dimensions together into a unified enterprise capability that delivers measurable value at the right scale and cost.