Indivar Nayyar is the Vice President of Data & AI at LTIMindtree, a global technology consulting and digital solutions company.
There is palpable excitement surrounding generative AI and agentic AI as organizations see the potential to boost revenues and reduce costs. Generative AI has already shown outcomes in areas of cognitive work like code generation, content creation and reasoning. These capabilities can be leveraged to completely reimagine workflows across various domains, whether in software engineering or marketing.
Beyond mere efficiency gains, AI possesses the transformative potential to reconfigure entire business models through the creation of products and services previously inconceivable with conventional AI approaches.
Warning: Hard Work Ahead
The opportunities presented by AI might be substantial. However, achieving meaningful outcomes with it remains far from straightforward. Most large enterprises operate on complex, layered technology infrastructures comprising legacy systems, evolving architectures and sprawling application stacks. The fundamental challenge lies in leading such organizations into an agentic AI world while delivering on the performance needs of today.
A Holistic AI Strategy, AI-Ready Data And Modern Infrastructure
Based on numerous implementations and conversations with customers, partners and practitioners, we have identified recurring patterns needed for establishing a robust foundation for AI implementation. Below are eight principles that enterprises should address in an agentic world.
1. From Integration Focus To Value Focus
For years, enterprises have invested billions in data initiatives—building massive extract, transform and load (ETL) pipelines and complex data lakes—often with underwhelming returns. It’s time to pivot: stop focusing on technical integration as the goal and start prioritizing business value.
Adopt a value-first approach to data and AI. Begin with specific use cases and avoid large-scale projects aimed solely at integrating data.
2. Shifting Culture From AI Use To AI Imagination
As foundational models and reasoning engines commoditize intelligence, imagination will matter more than raw intelligence. Instead of merely identifying use cases for AI, leaders should challenge teams to reimagine entire workflows and business models through new AI capabilities.
Build AI literacy and encourage curiosity—unlock your workforce’s potential to rethink the possible.
3. Treating Data As A Product, Not An Asset
The long-held belief that “data is an asset” has led to data hoarding—large ecosystems built without a clear purpose, often restricting access in the name of compliance. In an AI-powered world, data must be easily consumable, discoverable and purpose-driven.
Adopt a data-as-a-product mindset—the goal isn’t just delivering the right information, but driving the right value from it.
4. Embracing Data Fabric Over Centralized Data
Historically, enterprises consolidated data into a central data repository, presuming that it would ensure consistency and simplicity. However, the complexities of distributed data, M&A and divestitures, as well as the rise of unstructured data, further complicated the integration of data into a single platform.
Data fabric, on the other hand, emphasizes federated data and unified governance and security while allowing AI applications to access and process data where it resides. Focus on connecting data with a fabric pattern. A single, monolithic system can no longer effectively handle the diverse and evolving data needs of an AI-driven organization.
5. From Single Source Of Truth To Fit-For-Purpose Interpretation
The traditional concept of a single source of truth (SSOT)—a unified, consistent view of data—can fall short in an AI-driven world. Generative and agent-based AI models require context-specific interpretations of data to function effectively. Forcing all use cases into one definition strips away the nuance AI needs to operate accurately.
Organizations should embrace fit-for-purpose data interpretations—flexible data structures and semantics tailored to specific contexts and use cases. This doesn’t diminish the need for data quality; rather, it acknowledges that multiple valid “truths” can coexist, each serving a different AI objective.
6. From Tabular Model To Polyglot Persistence
Legacy data systems were built around tabular, relational models (like star schema or data vault). These are limiting for AI, which must work with varied data types—text, images, audio, graphs and more. Consider a polyglot persistence approach: choose the right database technology for each type of data and workload.
That means using NoSQL for semi-structured data, graph databases for relationships, vector databases for similarity search and so on. This approach enhances performance, scalability and coverage of diverse AI use cases.
7. Linear, Data-Centric To Accelerated, Observable AI-Centric Infrastructure
Today’s data infrastructure is optimized for analytics and structured data—not for AI agents or multimodal workloads. AI requires accelerated compute (GPUs, TPUs, LPUs), high-throughput storage and architectures built for training and inference at scale.
To manage this complexity, organizations need a single control plane—a unified platform to oversee compute, storage and networking across environments. This helps streamline operations, accelerate deployment and ensure governance.
8. Beyond Infrastructure-As-Code To Agentic InfraOps
Infrastructure-as-code has altered how we provision and manage infrastructure, enabling automation and repeatability. However, much more can be done with agentic InfraOps—AI agents with tool access—to autonomously manage infrastructure. This includes scaling resources, optimizing costs and even self-healing.
While still emerging, this paradigm promises a leap forward in operational efficiency and resilience.
Conclusion
We’re at a turning point as significant as the Industrial Revolution—only this time, cognitive machines, not mechanical ones, are reshaping work and society.
AI will likely transform every industry. The difference between success and failure will come down to mindset, skills and tools. To stay competitive, data and AI leaders—and the teams they lead—must rethink what’s possible and redesign their workflows for this new era.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
This post was created with our nice and easy submission form. Create your post!
