Agentic AI may be the future of autonomous enterprises, but getting there is anything but simple. “Are they accurate? Are they meeting the purpose?” These are the unanswered questions, according to Bala Prasad Peddigari, chief innovation officer for technology, software services business Group at TCS.
There are not many productionised systems currently available in the agentic AI space, he said while talking to AIM on the sidelines of the NIIT StackRoute Digital Architect Conclave.
However, Peddigari explained that one of the global trends currently driving the market is the push toward building autonomous enterprises, meaning “you want the enterprise to be much, much more autonomous, just like you have different private colleges going for autonomous status.”
“They have the ability to perform all their functions without depending on a university. Likewise, enterprises need to be autonomous in making decisions without relying on manual intervention,” he said.
Agentic AI, according to him, is providing this direction of thought, and as a result, many pilot engagements are underway.
“But what is not helping them is model accuracy. That demands two important elements: one is bringing contextual evidence into the retrieval augmented generation (RAG) model, and the second is fine-tuning that model with respect to the domain context, which is a very hard thing to do,” Peddigari shared.
Both elements need to be embedded as part of the model and require highly mature and deep technology skills. “That is currently lacking. We do not have it in masses, and that is not helping us scale faster,” he said.
“Even though in theory we can say we have so many—I gave you some from the agent marketplace: reflexive agents, purposeful agents, orchestration agents—these kinds of agents are available.”
“But are they accurate? Are they meeting the purpose? Are they addressing diversity? Are they adapting to change and meeting the need? That is still a question mark. These are things that are only tested over time,” Peddigari said.
Calling it an opportunity rather than a bottleneck, he said organisations are working towards it.
“And that requires more processing capabilities…and deeper skill readiness. If quantum processing becomes available, it will accelerate both decision-making and processing capabilities. Current capabilities are limited, but quantum could change the dynamics,” he said.
When all these elements converge, he added, meaningful progress will happen. And while the technology may be ready, he stressed that skill readiness and political will are equally crucial for real advancement.
State of Enterprises in India
According to Peddigari, Indian enterprises remain stuck at level one of AI maturity—focused largely on siloed experimentation without any real momentum toward institutionalising or democratising AI across the board.
In his assessment, more than 90% of Indian companies are currently dabbling with AI in fragmented ways, lacking both strategic cohesion and scalable frameworks.
The situation is even more dire for India’s micro, small, and medium enterprises (MSMEs), many of which haven’t even reached the experimental phase. For them, the challenge is not just technical but infrastructural and financial. “They do not even have the opportunity to experiment,” he said, emphasising that MSMEs urgently need sandbox environments where they can test their AI use cases without upfront capital commitments.
This, he believes, should be enabled through broader national frameworks such as the Viksit Bharat initiative or the Digital India program, bringing inclusivity into the AI growth story.
To address this, Peddigari stressed the importance of ecosystem participation. State-led innovation labs, like T-Works and T-Hub in Telangana, are a step in the right direction, but more needs to be done.
Providing access to GPU acceleration, AI-specific equipment, and model experimentation environments would help remove the uncertainty that currently deters MSMEs from investing in AI. Without confidence in tangible results or talent readiness, companies are unlikely to make those investments.
AI Adoption is Layered
Peddigari acknowledge the complexity of the landscape and breaks it into layers: start-ups, MSMEs, enterprise IT services firms and product-based companies. Startups benefit from free credits and sandbox offerings from major players like Google, Microsoft and NVIDIA, allowing them to experiment and rapidly demonstrate capabilities, even if they are not yet focused on scale.
MSMEs, meanwhile, want to benefit from AI but can’t afford failure. While some invest in talent and tools, many remain hesitant due to a lack of resources and skill depth.
Enterprise-grade service firms are trying to bridge this by forming partnerships with global MNCs. These collaborations allow them to scale internally through shared sandboxes and then deploy trained resources across client ecosystems.
Finally, large product companies and unicorns, which benefit from preferential rate cards and subsidised access to infrastructure, are better positioned to invest heavily in AI acceleration.
TCS’s Partnership
While companies like HCL have announced partnerships with major AI players such as OpenAI, TCS is approaching collaborations with a more integrated philosophy. Peddigari said the company focuses on building “360-degree relationships” rather than transactional tie-ups.
“OpenAI demands investments, that is not the way we work,” he said. “We want them to be part of our ecosystem, and we want to be part of theirs.”
He explained that this approach allows TCS to understand the assets, products, and solutions of a partner more deeply. “We learn that, we dogfood that, and then collectively take it to the market, tailoring it to the needs of the customer. That creates a win-win situation.”
Asked whether the company’s engagement with OpenAI was official, Peddigari clarified that it was “transactional at this moment”, possibly limited to a specific initiative. “I don’t have details to comment upon,” he added.
The post Accuracy and Precision Still a Distant Dream for Agentic AI? appeared first on Analytics India Magazine.
