TLDR
- Enterprises are moving past AI experimentation into disciplined deployment, with success rates higher than the 5% an MIT study reported, according to Vijay Guntur, CTO and head of ecosystems at HCLTech.
- Agentic AI, physical AI and edge devices are poised to become mainstream, reshaping workflows and industrial operations.
- Falling inference costs and global modernization needs will open trillion-dollar opportunities, though human change management remains the biggest barrier.
Enterprise AI is entering a new phase in which companies are getting more adept at deploying the technology not only to drive efficiency but to transform core services, according to Vijay Guntur, CTO and head of ecosystems at HCLTech.
“People have learned how to do it,” he said in an interview with The AI Innovator. “There’s been some frustration that getting things into production is not easy, and there’s a narrative around that. But we are also seeing lots more disciplined ways of getting things into production now.”
While AI adoption varies among companies, on the whole, “there are more enthusiastic enterprises today than there were in the past, and we believe this adoption rate is going to continue to grow,” added Guntur, whose company recently passed $100 million in AI-related revenue.
Despite a well-publicized MIT study saying that 95% of AI pilots fail, meaning only 5% succeed, his experience shows that “one in six or one in seven projects is succeeding today,” or a rate of 14% to 17%.
As for concerns about an AI bubble forming due to the billions of dollars being invested, Guntur said “people with deep pockets should not be worried. If they don’t invest, they will miss out for sure.”
He cited Google as saying it was “better off over-investing than under-investing.” If they do over-invest, it will just take longer to get their return on their investments. “Companies that have deep expertise, deep financial strength, they may over-invest a bit, but I think they’re on the right track of making sure their investment will pay off” by building services, he said.
Trillion-dollar market for AI
The acceleration of AI adoption is not globally uniform. North America, he said, is moving fastest, particularly in simulation, visual inspection and digital twins. Asia is also progressing quickly, with a mixture of experimentation and production systems.
Europe, he noted, remains slower but is heavily focused on modernization. “There’s a lot of legacy code, legacy systems sitting there,” he said. Europe’s need for upgrades is reflected worldwide. Guntur said the global modernization demand reflects a “trillion-dollar opportunity.”
Even emerging markets can get in on the action as they may be able to skip intermediate steps entirely. Drawing parallels to the mobile revolution, he said many countries avoided building landline infrastructure and jumped directly into smartphones. With AI, “you can skip generations of technologies,” he said, describing a potential leapfrog moment for developing economies.
Agentic AI, robotics, quantum outlook
“This year was a year of agentic but that, I think, will become mainstream,” Guntur said. He anticipates widespread adoption in 2026 as organizations incorporate agents into business processes, customer operations and software development.
The shift marks a break from early generative AI tools that focused on language generation. Agentic systems integrate reasoning, planning and multi-step execution – functions Guntur believes will reshape enterprise workflows.
He expects enterprises to launch “more product services with AI and start seeing more ROI” as these agentic capabilities mature and companies become more familiar with their use.
Guntur pointed to physical AI – AI-powered robotics and multimodal systems operating in real-world environments – as another major trend. Long considered futuristic, these technologies are now being deployed across ports, mines, manufacturing floors and product testing lines.
“We have demonstrated use cases – real business enterprises buying these solutions,” he said. “It’s real.”
He described rapid progress in computer vision and edge inferencing, which make it possible to evaluate materials, detect defects or track safety risks without human observers.
“Vision models are becoming real. Vision models are working; they’re affordable,” he said. Companies are integrating edge devices directly into industrial systems, creating self-monitoring infrastructure.
Humanoid robots – those shaped like people similar to Tesla’s Optimus – are still several years from widespread deployment, but he believes the trajectory mirrors advances in robotic surgery. Guntur predicts humanoid robots will be ready in three years or so.
New AI devices to mirror smartphone adoption
A further wave of transformation will come from edge AI – models running on-device rather than in the cloud – expanding across consumer and industrial hardware.
Smart glasses, drones and advanced cameras are early examples, but Guntur expects entirely new categories of AI-native devices to reshape the market. “OpenAI is already working on a device,” he said. “Google will launch, Meta will launch. Apple may also launch it,” he said. These devices could enable more responsive, context-aware AI experiences without heavy cloud dependence.
Affordability will drive adoption. If AI hardware drops below $100 or $200, he said, “it will proliferate,” allowing consumers to normalize AI usage globally. “If consumers get very comfortable with AI,” they’ll bring the same familiarity to work.
Inference costs to tumble
Looking further ahead, Guntur identified falling inference costs as the most important long-term driver of industry transformation. “Today, still, AI is not that affordable,” he said. However, Guntur forecasts that inference costs are going to drop “significantly” as advances in the AI stack continue.
Inference costs are the ongoing costs of using AI once a model is trained. Every time a user prompts an AI model and gets a response, inference costs are incurred, for example.
As inference costs fall, AI becomes cheaper to integrate into products, services and daily operations. Lower prices expand access, enabling “new applications, new products, new categories, new firms,” Guntur said.
He described a cycle where affordability drives innovation, which in turn accelerates disruption. Entire sectors could be reshaped not because large enterprises resist technology, he said, but because “they find change management very hard.” That will create openings for startups that are “going to go eat their lunch – 100% that will happen.”
As for quantum computing, while there have been advances of late, practical deployment remains years away. “It’s some time away, at least two years minimum,” he said, citing research at IBM, Microsoft, Google and Nvidia’s partner ecosystem. While promising, “it’s not ready for prime time yet,” and requires advances in qubit stability before commercial use is feasible.
Despite AI’s momentum, Guntur said the defining challenge of the next era will be human rather than technical. “Be confident, be bold, but take your people along,” he said. “That’s super-critical. Don’t underestimate change management in your journey. … Invest in your people.” He also encouraged leaders to experiment even if they risk failures. “Not everything will succeed, but the successes that you get are going to make up for” the failures.
