EY’s Raghunandan Pulijala explains how enterprises can turn analytics into business value, scale AI across Assurance and strengthen data governance.
Nikitha Sebastian31 Oct 2026
16:35IST
New Update
Enterprises continue to invest in data analytics and artificial intelligence (AI), but many struggle to translate these <a href="https://bitcomme.com/venture-capital-firm-octopus-ventures-leads-38-million-investment-in-braid/” title=”Venture capital firm Octopus Ventures leads $38 million investment in Braid”>investments into measurable business outcomes. Fragmented data environments, weak governance, organisational silos and limited adoption often prevent businesses from moving beyond dashboards and pilot projects. At the same time, AI is changing how Assurance teams assess risks, examine evidence and support critical decisions, while raising questions about human oversight and accountability.
Raghunandan Pulijala, Assurance Digital Leader, EY Global Delivery Services India LLP, discusses the challenges of translating analytics into business value, the role of AI in Assurance, the capabilities global capability centres (GCCs) need to build and how enterprises can balance faster access to insights with data quality and governance.
From insights to action
Enterprises continue to invest in data analytics, yet many struggle to translate insights into business decisions. Where does the disconnect typically occur?
The disconnect usually occurs between generating an insight and redesigning the decision it should improve. Many analytics programmes begin with platforms, dashboards or models rather than a clearly defined business question, an accountable decision-maker and a measurable outcome. Consequently, even strong analytics can remain advisory instead of becoming part of planning, pricing, risk management or operational workflows.
Weak data foundations add to the challenge. Inconsistent definitions, fragmented sources, limited lineage and unclear ownership reduce confidence in analytical outputs.
A second gap emerges at the organisational level. Analysts may focus on improving technical accuracy, while business teams need timely recommendations, explanations of uncertainty and clear courses of action. Skills and adoption also matter. When leaders and frontline users lack the capabilities or incentives to act on insights, established habits prevail.
Enterprises must design analytics around the entire decision loop. They should define the decision and the value at stake, assign accountability, establish trusted data and governance, integrate analytical outputs into workflows and track the actions that follow. Feedback on outcomes should then help refine the model. This approach moves analytics beyond reporting what happened to improving what the enterprise decides next.
How is AI changing the way Assurance teams assess risk, identify anomalies and support decision-making?
Artificial intelligence (AI) is expanding the scale, speed and precision of Assurance. It can ingest and compare large volumes of structured and unstructured evidence, identify unusual transactions, surface patterns across populations and refresh risk signals as data, controls and business conditions change. This allows teams to move beyond predominantly periodic and sample-based analysis towards broader coverage, earlier exception detection and more focused investigations.
Generative AI (GenAI) can also summarise evidence, connect relevant guidance, draft routine documentation and help professionals navigate complex information more efficiently. However, these capabilities support the Assurance process; they do not replace accountability for its conclusions. A flagged anomaly does not automatically indicate a misstatement, control failure or fraud. It prompts further inquiry, corroboration and professional scepticism.
The strongest model combines human leadership with technology. AI handles data-intensive analysis and prioritisation, while professionals evaluate context, challenge assumptions, determine the sufficiency and reliability of evidence and decide when to escalate an issue.
Used this way, AI can strengthen risk assessment and decision support while allowing Assurance teams to devote more time to areas that require professional judgement, sector knowledge and oversight.
As enterprises deploy AI across Assurance workflows, which decisions still require human judgement, and why?
Human judgement remains essential whenever a decision involves materiality, ambiguity, competing evidence or accountability to stakeholders. AI can identify patterns, rank exceptions and summarise information, but it cannot assume professional responsibility for the conclusion.
Assurance professionals must still determine whether evidence is sufficient and appropriate, whether an anomaly has a credible business explanation, whether a control issue is isolated or systemic and whether its severity warrants escalation. Judgement also plays a central role in assessing management estimates, interpreting complex transactions, evaluating bias and deciding how uncertainty should influence the Assurance response.
These decisions require business context, sector knowledge, ethical considerations and an understanding of consequences that data may not capture. Human oversight also matters when AI produces difficult-to-explain outputs, conflicts with other evidence or reflects limitations in the underlying data or model.
The objective should not be to place a person nominally in the loop. Enterprises must instead establish meaningful decision rights: what AI can recommend, what it can execute within defined guardrails and what requires review and approval from a responsible professional. Technology can strengthen judgement, but it cannot automate accountability.
What prevents organisations from scaling successful digital transformation pilots across business functions and geographies?
Pilots often succeed because they operate within narrow boundaries, receive strong sponsorship and avoid the complexity of the wider enterprise. Scaling exposes that complexity.
Different functions may follow incompatible processes, use different data definitions and operate on disparate technology stacks. Responsibility may also remain divided across business, technology, risk and local teams. A solution designed around the data, regulations or workflows of one market may require significant redesign before another market can adopt it.
Legacy platforms and weak interoperability create further friction. Moreover, short-term pilot funding rarely accounts for integration, controls, change management and ongoing product ownership.
The people dimension matters equally. If organisations do not change roles, incentives, skills and management routines, adoption remains local even when the technology works. Geographic expansion adds requirements around data residency, privacy, model risk and local regulations.
Enterprises need an operating model that supports scale rather than a collection of disconnected projects. This requires clear executive sponsorship, business ownership of outcomes, reusable data and technology components, common governance and global standards that allow necessary local variations. Funding should support end-to-end value streams rather than isolated use cases.
Finally, leaders must measure adoption and business value after deployment. This allows them to adapt, scale or discontinue solutions based on evidence rather than enthusiasm.
How should technology leaders measure the business value of data analytics beyond dashboard usage, reporting efficiency and operational productivity?
Technology leaders should begin with the business decision or value stream, not the dashboard. Usage, report cycle times and productivity serve as useful leading indicators, but they do not establish whether the enterprise made better decisions or realised business value.
A stronger scorecard should connect four levels: adoption inputs, workflow changes, operating outcomes and enterprise outcomes. Relevant measures include decision cycle times, forecast or risk assessment accuracy, error and loss reduction, control effectiveness, revenue conversion, margin improvement, working capital release, capital avoidance, customer retention and resilience.
The appropriate measures will vary by use case. However, the business leader who receives the benefit should own the outcome, while finance should validate the economic impact. Measurement also requires a credible baseline and, wherever possible, a counterfactual that establishes what would have happened without the analytics intervention.
Leaders must account for the full cost of the initiative, including data, platforms, engineering, governance, change management, security and ongoing model operations.
They should also review value over time because benefits often depend on adoption, process redesign and decision rights rather than technology performance alone. The central question is not whether people opened a dashboard, but whether an insight changed an action, improved an outcome and sustained that improvement.
Across banking, insurance, retail and manufacturing, what differences do you see in how organisations approach data-led transformation?
The overall direction remains similar across sectors, but each sector operates within a different decision-making context.
In banking, data-led transformation focuses heavily on risk-critical decisions involving credit, financial crime, capital, compliance and customer engagement. Explainability, auditability, model risk and resilience therefore set particularly high standards.
Insurance shares this regulatory intensity, but its transformation priorities centre on actuarial and claims data, underwriting, pricing, policy administration and long-duration liabilities. Legacy platforms and the need to improve customer interactions often make modernisation as important as advanced analytics.
Retail operates at a much higher frequency of customer and commercial decisions. Demand sensing, assortment planning, promotions, pricing, inventory, fulfilment and personalisation require organisations to connect consumer, product and supply chain data quickly across channels.
Manufacturing, meanwhile, focuses more heavily on assets and operations. Its priorities include predictive maintenance, quality, throughput, planning, safety, energy consumption and supply chain resilience. These initiatives combine enterprise data with sensor, engineering and operational data.
These differences do not represent maturity rankings. Leading organisations operate across all four sectors. The practical distinction lies in where business value concentrates, which data matters most and how much control an organisation requires before it automates a decision.
Sector context should therefore shape the architecture, governance and human oversight model from the outset.
As GCCs take on more complex digital and analytics work, what capabilities must they build to move beyond service delivery towards strategic business contributions?
To move beyond service delivery and make strategic contributions, GCCs must combine scale with ownership.
The first requirement is deep domain knowledge. Teams must understand the business decisions, risks and sector context behind their work rather than simply execute defined tasks.
The second is product and platform expertise. This includes product management, data engineering, advanced analytics, AI, architecture and lifecycle governance. These capabilities allow GCCs to co-create reusable solutions, operate them reliably and improve them based on user outcomes.
Third, GCCs must strengthen their business partnering capabilities. Leaders should shape priorities alongside global functions, build business cases, influence investment decisions and translate technology initiatives into measurable business value.
End-to-end accountability also matters. GCCs must take responsibility for quality, adoption, controls, cybersecurity, responsible AI use and benefits realisation.
In India, the depth of technology and domain talent creates a strong opportunity. However, GCCs must support this talent with clear career paths, continuous learning, mobility and leaders who can work across cultures and geographies.
Finally, GCCs need an innovation operating model with clear sponsorship, funding, prioritisation and pathways to scale. They can make this transition when global teams involve them early in strategy, solution design and product decisions and hold them accountable for outcomes rather than activity alone.
With enterprises operating increasingly complex data environments, how should leaders balance faster access to insights with data quality, governance and accountability?
Leaders should design speed and governance together. The objective is not unrestricted access to every dataset, but faster access to trusted, appropriately permitted data that supports specific business decisions.
A governed self-service model can provide curated data products through a catalogue, with common business definitions, documented lineage, quality indicators and role-based access.
Clear ownership forms the foundation of this approach. Business data owners should define meaning and acceptable use, stewards should maintain quality, technology teams should manage platforms and controls, and risk or Assurance functions should provide independent challenge where necessary.
Pre-approved access pathways can accelerate routine requests when organisations have already established consent, purpose and controls. Higher-risk use cases can follow additional review processes.
Quality controls should operate throughout the data lifecycle. These include profiling, validation, reconciliation, change management and monitoring for deterioration. The same principle applies to AI: models and agents should use approved data, preserve traceability and operate within defined decision rights. Organisations should also establish escalation mechanisms when confidence or data quality falls below agreed thresholds.
This approach makes governance an enabler of speed. Leaders gain faster access to insights because organisations resolve definitions, permissions and accountability in advance, rather than renegotiating them for every request or discovering gaps after making a decision.
