New Delhi, July 13 -- Digitide Solutions' Chief Strategy, Solutions & AI Officer says enterprises that redesign operating processes, rather than simply deploy AI tools, will emerge as long-term winners.
As enterprises move beyond AI experimentation, the conversation is shifting from model performance to governance, accountability and operational redesign. Organisations are increasingly recognising that scaling AI requires far more than better algorithms. It demands rethinking business processes, defining clear ownership and embedding human oversight into critical decisions. In an exclusive conversation with TechCircle, Sandeep Malhotra, Chief Strategy, Solutions & AI Officer at Digitide Solutions, a technology services and business process management company, discusses why workflow redesign has become the biggest hurdle to AI adoption, where enterprises are already seeing measurable business outcomes, and why governance, rather than AI models, will define the next phase of enterprise AI. Edited excerpts from the conversation. Q. Has the enterprise AI conversation shifted from "What can AI do?" to "What should AI not do?" Where are companies drawing that line today? Malhotra: Absolutely. Early discussions focused on what AI could automate or accelerate. Today, enterprises are asking where AI should not operate autonomously. The line is increasingly being drawn around decisions with regulatory, financial, ethical or reputational implications. In areas such as underwriting, claims rejection, fraud escalation, pricing and customer dispute resolution, AI can analyse information and recommend actions, but the final decision still requires human judgment. The industry is steadily moving towards a human-in-the-loop model. AI is expected to augment critical decisions, while accountability remains firmly with people.
Q. If human oversight remains essential, are enterprises overestimating the cost savings AI can deliver?
Malhotra: In some cases, yes. Many organisations still view AI primarily as a labour substitution tool, which creates unrealistic expectations around immediate cost savings. The biggest gains come from redesigning workflows so that humans and AI work together more effectively. Human oversight is not a weakness, it is a design principle, particularly in industries such as insurance and healthcare where trust and regulatory compliance are critical. AI can significantly reduce manual effort, improve turnaround times and increase decision accuracy. However, enterprises must also consider the costs of model training, data preparation and ongoing operations before calculating returns.
Q. What is the biggest reason AI pilots fail to scale: technology limitations, poor data or the absence of workflow redesign?
Malhotra: Workflow redesign remains the biggest obstacle. Many AI pilots succeed because they solve a narrowly defined problem. Scaling them, however, requires integrating AI into core business processes with clear ownership, governance, exception handling and measurable outcomes. Poor data and legacy technology certainly slow progress. But the larger issue is that organisations often try to insert AI into outdated operating models instead of redesigning those processes around AI-enabled decision-making.
Q. In insurance, where is AI creating the most measurable business impact today: claims, underwriting, fraud detection or customer servicing?
Malhotra: Claims processing and fraud detection are currently delivering the fastest and most visible business outcomes. Claims processing is highly document-intensive and time-sensitive, making it well suited for AI-driven document understanding, intelligent routing, anomaly detection and faster settlements. These improvements enhance operational efficiency while improving customer experience. Fraud detection is another high-impact area because AI can identify behavioural patterns that conventional rule-based systems often miss. AI-powered claims auditing is also gaining traction among insurers. Underwriting continues to benefit from improved risk segmentation, while customer service is becoming more personalised through AI-driven interactions.
Q. Many organisations are investing heavily in AI models. Are they underinvesting in governance, monitoring and auditability?
Malhotra: In many cases, yes. Most investment has gone into improving models and accelerating experimentation, while governance, monitoring and auditability have received comparatively less attention. As AI becomes embedded in core business processes, organisations must continuously monitor model drift, explain AI-generated decisions, detect bias and maintain accountability over time. For regulated sectors such as insurance, healthcare and banking, governance is no longer just about compliance. It is essential for building trustworthy, scalable AI. Ethical AI practices, explainability, high-quality data and audit controls are rapidly becoming baseline requirements.
Q. As AI becomes embedded in core business processes, who ultimately owns accountability for AI-driven decisions: the technology team or the business?
Malhotra: Accountability ultimately rests with the business. Technology teams are responsible for building reliable models, maintaining infrastructure and implementing safeguards. Business teams, however, own decision policies, operational outcomes and customer impact. AI should never become an accountability grey zone where difficult decisions are attributed to "the system". Successful organisations will establish shared governance, but ownership must remain clearly with the business wherever AI influences operational or customer outcomes.
Q. Why are more enterprises moving from building AI capabilities internally to ecosystem-led partnerships?
Malhotra: Enterprise AI has become far more complex than simply building models. Successful AI programmes require domain expertise, quality data, workflow integration, governance frameworks and scalable platforms. Few organisations possess all these capabilities internally at scale. In-house expertise remains strategically important, but ecosystem partnerships accelerate deployment, reduce execution risk and provide specialised capabilities that would otherwise take years to develop. That is particularly relevant in sectors such as insurance, where AI success depends as much on operational integration and regulatory understanding as it does on technical capability.
Q. Two to three years from now, what will distinguish AI leaders from laggards: access to better models, better data or better operating processes?
Malhotra: Better operating processes will be the biggest differentiator. AI models will become increasingly accessible, and data quality will remain important, but sustainable advantage will come from how effectively organisations redesign their operations around AI. The leaders will connect data, workflows, governance and decision-making into a single operating system. They will not simply deploy AI tools, they will embed AI into how work gets done and how outcomes are measured. Laggards, meanwhile, will continue pursuing isolated pilots and fragmented use cases. The next phase of enterprise AI will be defined less by access to the best models and more by the discipline to operationalise AI consistently across the business.
Published by HT Digital Content Services with permission from TechCircle.