Why Data May Be the Biggest Challenge for AI Agents

Photo By: Mohamed Nohassi

Businesses are investing heavily in AI agents, attracted by the promise of automating tasks, improving productivity, and streamlining operations. Unlike traditional AI tools that primarily generate content or answer questions, AI agents are designed to take action, complete workflows, and interact with multiple business systems.

As interest in the technology grows, attention is increasingly shifting to a less visible but critical issue: data.

For many organizations, the success of AI agents depends not only on the capabilities of the underlying models but also on the quality, accessibility, and governance of the data those agents rely on.

The challenge is one that industry leaders are increasingly focused on. Harsha Kumar, CEO of NewRocket, leads a company that helps enterprises deploy AI, automation, and digital workflow solutions. As organizations pursue agentic AI initiatives, the ability to connect, govern, and trust enterprise data is becoming a critical factor in determining whether those investments deliver measurable business value.

AI Agents Depend on Reliable Information

AI agents are designed to retrieve information, make decisions, and perform tasks based on the data available to them. To do that effectively, they need access to accurate and current information across an organization.

Many enterprises, however, operate with data spread across multiple applications, databases, and cloud environments. Information may exist in different formats, follow different standards, or be managed by separate teams.

When data is fragmented or difficult to access, AI agents may struggle to provide consistent results. Even highly capable models can be limited by incomplete, outdated, or conflicting information.

This has led many organizations to recognize that data readiness is an important part of any AI strategy.

The Cost of Getting It Wrong

The stakes are growing as organizations accelerate their AI investments. Gartner predicts that by 2028, 40% of agentic AI projects will be cancelled because of escalating costs, unclear business value, or inadequate risk controls.

While organizations often focus on the capabilities of AI models, the underlying data foundation can play a significant role in determining whether projects deliver measurable outcomes. When data is not trusted, connected, or easily accessible, AI agents may struggle to operate effectively.

The consequences can extend beyond technical performance. Decision-making can slow down, customer experiences can become fragmented, and organizations may find it difficult to demonstrate a return on investment. In these situations, the efficiencies AI is intended to create can be diminished by the complexity of the data environment supporting it.

Governance Is Moving to the Forefront

As companies explore the use of AI agents in business processes, governance is becoming an increasingly important consideration.

Organizations must determine what information agents can access, what actions they can take, and how those actions are monitored. These questions become particularly important when agents interact with sensitive business data or operate in regulated industries.

Without clear governance policies, companies may face challenges related to security, compliance, and accountability.

As a result, many organizations are evaluating not only the capabilities of AI systems but also the controls needed to manage them responsibly.

Data Preparation Can Be a Significant Undertaking

Deploying AI agents often requires more than connecting a model to existing systems.

Organizations may need to integrate data sources, improve data quality, establish access controls, and create processes for ongoing monitoring and management. In some cases, these efforts can represent a substantial portion of the work involved in preparing for large-scale AI adoption.

This reality is prompting many business leaders to view data infrastructure as a foundational requirement rather than a separate initiative.

Building the Foundation for Agentic AI

The rapid development of AI agents has generated significant excitement across industries. Their ability to automate complex tasks and support decision-making has the potential to reshape how work gets done.

At the same time, the effectiveness of these systems remains closely tied to the data environments in which they operate.

As organizations move from experimentation to implementation, many are finding that success requires more than advanced AI models. It also requires reliable, accessible, and well-governed data.

For enterprises pursuing agentic AI, the challenge may not be whether the technology is capable enough. It may be whether the underlying data foundation is ready to support it.

About Harsha Kumar

Harsha Kumar, CEO, NewRocket

Harsha Kumar is the CEO at NewRocket, helping elevate enterprises with AI they can trust, leveraging NewRocket’s Agentic AI IP and the ServiceNow AI platform.

He is an Entrepreneur, Technologist, and Growth-stage CEO with a proven track record of driving significant revenue and EBITDA growth, customer satisfaction and career growth for team members, across software platform, product and services companies.

He led Prodapt as President and then CEO from 2016-2024, a period of high growth (7x in 8 years), and established Prodapt as a leading technology transformation and RPA partner for Telcos. In 2023, Gartner recognized Prodapt as an industry leader across America, Europe, and LATAM.  Raised PE from Affirma Capital and received strategic investment from ServiceNow as their strategic Telco industry partner.

He’s known for his hands-on and people-centric leadership approach. He enjoys building businesses by deploying GTM approaches that beat larger more established competitors.  He is accessible, consistent and fair, persistent, supportive yet demanding, empowers his teams and holds them accountable for results. He’s immensely proud of his team members that have gone on to lead successful and innovative businesses.

Harsha played a number of senior leadership roles to establish and grow Virtusa into a leading IT consulting and transformation services firm.  He set up advanced technology centers, started and led BUs for Telecom, Media & Information, Insurance, Healthcare, and Technology industries.  He was a member of Virtusa’s core leadership team as it became a leader in consultative transformation of enterprise workflows (Pega Elite partner) and scaled from a small $13M private company to a $1B+ publicly listed company.

Earlier, Harsha co-founded EC Cubed – a B2B software platform company and raised significant venture capital.  As its CTO, he architected and developed a workflow engine and other components.  He started out at Bellcore, developing telecom B/OSS products.

Harsha received an MS from the University of Maryland working with Hughes Network Systems under the guidance of the legendary Prof. Ben Shneiderman, and a B.Tech from IIT Delhi, where he was awarded the prestigious outstanding student leader award. He is a co-author of “Enterprise E-Commerce” which has been a textbook at leading US universities. He has completed exec programs at ISB and Stanford GSB.

Residing near NYC, he is family oriented and an ardent fan of his kids’ activities.  He likes to sing, run, play racquet sports and an occasional round of double bogey Golf.