Data-driven Approach while Rethinking Businesses with AI

What strategies should businesses adopt to build a foundational AI strategy that maximizes value from data initiatives? 

Organizations must adopt a data-driven approach while strategically rethinking their businesses with AI. This involves establishing a solid data and AI strategy, integrating AI into business processes, fostering a data-driven culture, and prioritizing governance and compliance. However, given the complex nature of enterprise data ecosystems, unifying all types of data into a single lake is not practical or viable. Therefore, to effectively harness all types of data's potential, and to ensure data quality, accuracy, completeness and consistency across multiple data sources requires a comprehensive approach encompassing data strategy, data governance, integration, validation, robust data quality checks, security, trust, ethics, privacy and compliance. Modern data architectures emphasize data organization and fingerprinting for efficient access and AI model training. This enables access control, synthetic data versioning, and data security measures while accelerating AI development and improving model accuracy and reliability.

AI initiatives thrive on a unified data strategy that leverages AI innovations to drive real business value. It's crucial to consider scalability when designing and building AI solutions to handle increasing data volumes and complexity. A unified data strategy provides a clear roadmap to manage/ govern data on all the required data capability components, envisioning it through business value relevance and ensuring it through a defined funding model to implement and run it sustainably.  A data strategy roadmap can help to comprehensively address aspects of people (operating model), process (data capability processes) and technology (data and technology architecture) thereby making it a foundation for building successful AI initiatives.

How is a data-centric approach enabling companies to reshape traditional business models, and what are some best practices for becoming AI-native? 

In order to protect the business and stay competitive, businesses need to leverage more data to generate more micro insights to create arbitrage and competitive opportunities. It would also need to leverage AI to create a faster time to market for its products and services, to improve and compete on standard business metrics, like net promoter score, wallet share, operational excellence, negotiated cost of goods, optimized supply chain, etc. Firms embracing the modern data estate enables them to remain competitive in the AI world.

AI enables new ideas to create new products and services for companies to venture into adjacent markets. AI through faster master analysis, and lower cost of entry to creating an AI-enabled product or service, enables the breakdown of the vertical silo. Companies that embrace new market innovation are likely to remain competitive while others may face disruptive forces.

Additionally, with AI, the cost of delivering products and services is reduced. A competitive business can dramatically play in the price band as its COGS (Cost of Goods Sold) becomes quite low to afford a wide arbitrage model. However, that would also mean that it must disrupt its own legacy business else it will face lower entry disruptors challenging their existing footprint.

What initial tech and data investments should companies focus on to drive productivity and responsibly harness AI's potential?

Legacy technology and outdated processes can hinder the successful implementation of AI-driven data strategies. Heterogenous infrastructure, technology debt, outdated hardware, complex business processes with significant human interventions, lack of end-to-end business process visibility, and insufficient data from these processes are some of the common obstacles. 

As enterprises identify high-value AI use-cases, to overcome challenges and harness the full potential of AI, organizations should adopt an AI-first transformation approach built on a cloud foundation. To further capitalize on investments, enterprises must secure data by investing in open data fabric, agentic AI orchestration and knowledge harnessing. This approach establishes a resilient and secure infrastructure, streamlines business processes, and enhances data visibility, enabling effective scaling of AI initiatives. To ensure success, enterprises must prioritize data governance, privacy and security, fostering a data-driven culture, and leveraging data fingerprinting techniques. A data-driven culture empowers employees to make informed decisions based on data insights, while data fingerprinting techniques enable organizations to untangle complex data ecosystems and optimize data management.

What key challenges and trends will shape the data analytics landscape in 2025, particularly around generative AI and the creation of AI ecosystems? 

The upcoming year will be a complex interplay of challenges and trends. Multi-modal data management where all data, structured docs, videos, images, communications, conversations, and personal information both inside the enterprise and the knowledge of the crowd can be accessed, integrated and consumed for AI. This is a challenge and an opportunity for innovation. 

The "cost of a token" is exploding as more and more language models, large and narrow, come into the AI ecosystem. If unchecked the cost of "getting an AI to answer" or doing an "AI Action" will make ROI questionable. A better design and understanding would need to be put in place to attribute, allocate, forecast and optimize the AI cost. A "FinOps for AI" will be needed that understands and regulates the "Infrastructure of AI" which consists of data assets, compute assets, I/O (Input/Output) assets, network assets, neural assets etc.

Moreover, responsible AI is a key need. To make sure all data that feeds AI comes from unbiased sources that are corroborated with journalistic discipline. The new data governance paradigm redefines "data quality" by using source reputation systems, multi-source validations and other techniques that apply to unstructured data. A lineage of knowledge which is more advanced than creating data lineage of structured data would also need to be in place.

What best practices ensure data quality, governance, and responsible AI deployment for trustworthy, data-driven decision-making? 

As organizations evolve from being data-driven to embracing an AI-first strategy, data continues to be the essential fuel for their AI engines, making its quality & governance more crucial than ever. Governing data in the AI world comes with its own set of challenges such as lack of trusted sourcing, anomalies & bias in data, the presence of large unlabeled structured and unstructured data along with aggravated privacy & security concerns.

To manage these challenges organizations should follow below mentioned best practices -  

1. Data & AI COE – Enterprises should establish a dedicated data & AI governance CoE with stakeholders from legal, privacy, business, AI model & data owners to govern, publish policies, and guidelines & manage the AI operations 

2. Data Fingerprinting – Organizations should fingerprint their extensive data footprint through the creation of data/model catalogues, data lineage, classification, data dictionary, etc. This approach will enhance overall governance by ensuring that data and model aspects are documented, tracked, traced, and evaluated, fostering trust in the AI systems. 

3. Data Quality Monitoring – Organizations must establish controls over data & AI, including profiling, data remediation, bias management, training data labelling, and data annotations, to ensure that high-quality training, validation & test data are used by the models. 

4. Data Privacy & Compliance Controls – Subsequently, controls such as personal data identification, personal data protection, consent management, and compliance assessments concerning various regulations (GDPR, EU AI Act, CPRA, etc.) including risk assessments, etc. should be implemented to ensure privacy and compliance in governance.  

5. Data Security Controls – Create a robust security layer over the AI systems to protect data from attacks such as data poisoning, inferencing, prompt injections, etc. through context-based access, security guardrails, etc. in addition to the standard security controls 

6. Data Discovery & Consumption – Consumption of data & AI through a centralized marketplace for easy discovery & responsible distribution across the organization.

Infosys has been committed to driving AI-first for business growth in a responsible manner and is among the first IT services companies globally to receive the ISO 42001:2023 certification for Artificial Intelligence Management Systems.

Read More:

Transforming 2025 Driven by AI, Machine Learning, and Robotics