Data readiness: Your first step to agentic AI success

At NTT DATA, I’ve seen firsthand how AI — especially GenAI in recent years, and now also agentic AI — can revolutionize organizations in every industry.

Autonomous agentic AI systems reason, plan and act to achieve goals, often in coordination with other agents. However, these systems only deliver the intended ROI if they have access to large volumes of high-quality, context-rich and real-time data.

This means you can only access the true power of agentic AI once you understand and achieve data readiness. As a global IT leader aptly put it at an event I attended recently, “It’s the combination of data and AI that will really take you places.”

The cornerstone of agentic AI

In NTT DATA’s Global GenAI Report, only 53% of organizations said they had already addressed data readiness while assessing GenAI-related business opportunities — but 95% expected to have done so within the following year, showing a sharply rising awareness of its importance. Respondents also said data readiness was the top lesson they had learned from their previous GenAI deployments.

As much as these statistics and the global IT leader’s statement are true for GenAI, they resonate even more deeply in the context of agentic AI, where data becomes the backbone that supports the full ecosystem.

Consider a scenario where a multinational drinks manufacturer wants to implement agentic AI to streamline its supply chain. The system would need to access and process vast amounts of data from various sources, including sensors, logistics providers and market analysts. If the data is not properly cleaned, integrated and governed, the system will make suboptimal decisions.

The undesired outcome would not be due to a lack of technological prowess or strategic business planning but simply result from a lack of data readiness.

What is data readiness?

Focusing on data readiness before deploying AI is like doing the prep work for baking a cake. You may have bought all the ingredients, but if they’re not of good quality, neatly measured or mixed well, your cake is likely to flop.

In the world of GenAI, this means having first-rate, well-organized data that’s clean and relevant so the AI can whip up creative content without any hiccups. The better the data, the more impressive the performance of your GenAI engine.

Agentic AI, in particular, needs dynamic, multisource and multiformat data streams to reason and act. This data should be contextually rich, semantically interoperable (not only shared but also understood by different platforms or applications) and available in real time.

Moreover, multiagent systems need synchronized data environments to enable collaboration, delegation and negotiation among agents.

In a multiagent system, different agents may be responsible for different aspects of the supply chain, such as procurement, inventory management and logistics. For the system to work effectively, the data from each of these agents needs to be synchronized and integrated in real time, allowing the agents to collaborate and make decisions based on a shared understanding of the supply chain.

• ALSO READ → NTT DATA and Microsoft simplify multiagent AI systems at scale

6 steps to data readiness for agentic AI

Achieving data readiness for agentic AI can be broken down into six key steps:

1. Understand your decision-making process

Clarifying the data you need to make business decisions is crucial, regardless of the technology you’re using to deliver and act on insights. This involves understanding business processes, identifying the key decisions to be made and determining what data you need to support those decisions.

2. Check data quality and integrity

The importance of structured, labeled and trustworthy data cannot be overstated. Here, human-in-the-loop validation is essential, as it helps to ensure that the data is accurate and reliable.

3. Plan data ingestion and integration

You’ll need real-time pipelines and application programming interfaces to continuously feed agents. AI-driven “extract, transform, load” (ETL) tools — which extract data from multiple sources, transform it into a standardized format and load it into a target system — can automate ingestion from legacy systems, making data integration easier and cheaper.

4. Set data migration in motion

Moving from siloed systems to cloud-native, agent-ready architectures is vital. Process mining — analyzing and improving business processes by extracting knowledge from event logs recorded by your systems — can help you identify and prioritize migration targets, allowing you to migrate the most critical data first.

5. Tackle data coding and semantics

The ability of AI agents to reason depends on ontologies, taxonomies and metadata tagging. These tools are all used to organize and interpret information, but they serve different purposes.

Taxonomies group items into hierarchical categories and subcategories based on their shared characteristics. For example, a taxonomy for a retailer might categorize products into electronics, clothing and home goods.

Ontologies go a step further by defining the relationships between those categories and by providing a richer structure that describes how different concepts interact with one another. For instance, an ontology might define the relationship between a product and its manufacturer, or between a customer and their order history.

Then, metadata provides information about the content, context and structure of data, such as the creator, the date it was created and the file type.

Think of an AI agent that has to recommend products to customers based on their browsing history. The agent needs to be able to understand the relationships between different products, customers and their preferences. Ontologies, taxonomies and metadata tagging allow it to make sense of this complex data and provide accurate recommendations.

6. Maintain data governance and compliance

Ensuring that your AI agents operate within regulatory and ethical boundaries is critical. Implement audit trails, access controls and explainability frameworks to check that your data is being used responsibly and complies with relevant regulations in your industry.

• ALSO READ → The AI responsibility crisis: why executive leadership must act now

How AI can accelerate your efforts

AI itself — in various applications — can play a big role in accelerating your data readiness. For instance, it can be used for data profiling, to scan and classify data and to assess data quality. It can automate deduplication, normalization and contextual tagging.

Large language models can autoclassify and enrich your data with additional context and meaning, making it easier to integrate and analyze data from different sources.

AI can also accelerate traditional ETL processes and monitor data usage while flagging anomalies and enforcing policies autonomously.

At NTT DATA, we’ve developed several accelerators that can help you speed up your data-readiness journey. These include AI-powered data profiling and cleansing tools, as well as data governance frameworks that can help you keep your data accurate, reliable and compliant with regulations.

Common challenges to overcome

Despite the clear benefits, you may face several challenges in achieving data readiness.

To get started, you can use tools like maturity models and ROI calculators to assess your current data readiness. You can also seek the advice of expert partners with experience in implementing not only agentic AI systems, but also the wider technological ecosystems in which they operate.

Data silos can be addressed using federated learning (a machine-learning technique that enables AI model training on decentralized data) and data-mesh strategies (with data ownership and responsibility distributed across teams in your organization, allowing for more agile data management).

If you have a legacy system that is incompatible with modern data architectures, you can use an AI wrapper to extract data from that system and integrate it with other data sources. This unlocks the value of the data without having to replace the entire system.

Then, change management remains essential. Align your business and IT teams through maturity assessments and pilot programs.

Are you ready?

Data readiness is a strategic imperative that will define the success of your AI initiatives.

By understanding what data readiness entails and taking concrete steps to achieve it, you can position AI, GenAI, and especially agentic AI as the driving force behind your organization’s digital transformation.

Are you ready to embark on this journey? We’re here to support you every step of the way.

• ALSO READ → A force for good: How AI and GenAI are reshaping our world