Why are companies struggling to turn AI experiments into actual business results?

Companies are investing heavily in AI technologies while struggling to translate pilot projects into business value. Data readiness, ownership, and trust raise challenges in effectively implementing AI solutions. Organizations must redefine existi...

Agencies
Representative image
Artificial intelligence has moved rapidly from research laboratories into boardrooms, offices and everyday business operations. Companies across industries are investing heavily in AI tools, experimenting with chatbots, deploying AI agents and exploring ways to automate routine tasks. Yet, despite the growing enthusiasm and spending, many organisations are struggling to translate these experiments into measurable business results.

The problem is not necessarily a lack of technology. Instead, companies are discovering that successfully testing an AI tool and making it work across an entire business are two very different challenges. A pilot project may deliver impressive results in a controlled environment. However, when the same technology is introduced into everyday operations, it often encounters fragmented data, outdated processes, regulatory requirements and resistance to change. In many cases, organisations also struggle to establish who is responsible for turning an AI experiment into a business outcome.

Rohit Kedia, CEO of Xoriant, a digital engineering services company specializing in building and deploying AI systems, believes this growing disconnect reflects a fundamental change in how businesses need to approach technology. "For the first time in the history of digital engineering, we have moved from a deterministic realm to a probabilistic one. Every enterprise system we have built over the last few decades was designed to behave the same way every time. You gave it an input, you knew the output, and you could test, govern and budget for it accordingly. AI doesn't work like that," Kedia said.


He believes many companies are still attempting to manage AI using traditional technology practices, even though the technology operates differently. "Yet most enterprises are still trying to manage probabilistic technology with deterministic-era habits. That is not a technology problem but a human one," he added.

Why do most AI pilot projects fail to deliver business value?

Kedia identifies four major reasons behind the gap between AI pilots and business value. The first is data readiness. An AI application that performs well in a controlled environment may struggle when exposed to fragmented and unstructured business information.
ADVERTISEMENT

The second challenge is ownership. Technology teams often lead AI pilot projects, but the actual business outcomes belong to other departments. When responsibility is divided, nobody may be fully accountable for taking the project forward.

The third issue is trust and governance. Businesses operating in regulated sectors such as banking, financial services and healthcare need to validate AI-generated outputs, monitor their performance and explain how certain decisions are reached.

The fourth challenge is the absence of human judgment in redesigning business processes.

According to Kedia, companies frequently introduce AI into existing workflows without questioning whether those processes should be changed in the first place. "AI can accelerate a process, but only people can decide which process deserves to exist, where AI should act, and what outcome it should serve," he said.
ADVERTISEMENT

Companies are measuring AI success using the wrong metrics

Another problem is the tendency to measure AI adoption through technical achievements rather than business performance. The number of AI tools deployed, tasks automated or employees using AI may indicate growing adoption. However, these numbers do not necessarily show whether the technology is improving revenue, reducing operating costs or delivering better customer experiences.
ADVERTISEMENT

Kedia believes successful AI implementation should have three characteristics. It should be measured through a business metric rather than a technical one, become part of an actual workflow instead of operating alongside it, and have a clearly identified human owner.

Why human judgment remains important as AI becomes more capable

The growing adoption of AI agents has raised questions about how much work businesses should delegate to machines and which responsibilities should remain with employees. While AI can process information, identify patterns and execute repetitive tasks at considerable speed, business decisions often require industry knowledge, experience and an understanding of circumstances that may not be captured in data.

Kedia believes human involvement should not be treated as a final checkpoint in an otherwise automated process. Instead, human ingenuity and artificial intelligence need to work together, with each complementing the other's strengths.

Kedia said the industry is moving towards a more outcome-driven talent model. This includes Forward Deployed Engineers, who understand client requirements and industry challenges; Platform Development Engineers, who convert solutions into reusable capabilities; and AI Full-Stack Engineers, who work across the development lifecycle.

The shift suggests that technical expertise alone may no longer be sufficient. Understanding the business problem and identifying where technology can make a meaningful difference are becoming equally important.

What should companies do to make AI investments deliver results?

For companies struggling to move beyond AI experiments, the next step may not be adopting another model or purchasing another tool. It may require rethinking how work is organised. Kedia believes businesses must begin with a clearly defined business problem rather than selecting an AI technology and searching for ways to use it.

They also need to redesign existing processes instead of simply adding AI to outdated workflows. Investments in data quality, governance, continuous monitoring and accountability are equally important. AI systems also cannot be treated as one-time deployments. Their performance needs to be evaluated and improved continuously, particularly when they are used in business-critical operations.

"The next phase will be defined less by which models an organisation uses and more by how intelligently it applies them," Kedia said.

He expects AI to become increasingly integrated into enterprise platforms, processes and decision-making systems, with human employees and AI agents working together to deliver specific outcomes.

As more companies gain access to similar technology, the ability to apply it to real business problems, build trust and redesign workflows will matter much more. As Kedia puts it, the real competitive advantage will belong to organisations that close the gap between AI innovation and its application, with human ingenuity leading and AI amplifying it.
Download
The Economic Times Business News App
for the Latest News in Business, Sensex, Stock Market Updates & More.
Download
The Economic Times News App
for Quarterly Results, Latest News in ITR, Business, Share Market, Live Sensex News & More.
READ MORE
ADVERTISEMENT

READ MORE:

LOGIN & CLAIM

50 TIMESPOINTS

More from our Partners

Loading next story
Business News › News › Company › Corporate Trends › Why are companies struggling to turn AI experiments into actual business results?
Text Size:AAA
Success
This article has been saved

*

+