More AI Agents Than Humans? Experts On Jobs, Degrees And Guardrails

Somewhere in India, a teenager is staring at two screens. On one screen is an offer letter from a prestigious university - a four-year computer science degree that will cost their family dearly. On the other, a swarm of AI agents is drafting a business plan, building a prototype and already writing to potential clients. That was the picture futurist Mike Walsh painted at the NDTV World Summit 2026, as he predicted that AI agents could soon outnumber people and that young Indians may hire their first digital employee before they land their first job.

It is a striking vision. But experts say the reality is more cautious, and more complicated.

"The most interesting development is that there could soon be more AI agents on the planet than people. We are already seeing the early signs through tools such as ChatGPT and other AI assistants, and India is likely to be one of the most important markets for how this transformation unfolds," Walsh said at the summit.

READ | AI Agents To Outnumber Humans Soon? What Futurist Mike Walsh Told NDTV

Dr Srinivas Padmanabhuni, a PhD in Artificial Intelligence (AI) from University of Alberta, Edmonton, Canada and the CTO of AiEnsured agrees with Walsh's direction. He cites Nvidia as an example to explain where this is heading. Nvidia CEO Jensen Huang has said the company expects to run around 7.5 million AI agents alongside roughly 75,000 human employees over the next decade, about 100 agents for every person.

On what counts as an AI agent Padmanabhuni told NDTV: "Unlike reactive assistants, agents are proactive, stateful digital workers that use APIs. They autonomously execute multi-step goals, such as managing vendor invoices or making micro-transactions, without needing manual human approval for every decision." 

So, how far are we from autonomy?

"We are in supervised autonomy, much like Level 2 self-driving cars. McKinsey case studies highlight agents automating complex workflows, but errors compound across multi-step tasks, so human-in-the-loop guardrails and liability governance are still needed," Padmanabhuni said. Level 2 autonomy, which he cited, is partial driving automation where the vehicle controls steering and speed together, but the driver must keep watching the road and be ready to take over at any moment.

So, where does it first land in India?

According to Padmanabhuni, customer service and back-office work lead, followed by coding, with personal errands last, because structured enterprise settings let agent "swarms" scale before individuals trust agents with money. 

"The BPO (business process outsourcing) and KPO (knowledge process outsourcing) sectors are already shifting from headcount billing to Agentic Process Outsourcing (APO) and human-in-the-loop orchestration. High-volume, analytical tasks are handed to agent swarms while humans become managers," he highlighted.

Walsh also said AI agents would soon be able to make low-value payments without requiring the user's approval every time, laying the foundation for "an entire ecosystem of autonomous personal AI."

Where To Draw The Line 

While Dr Padmanabhuni puts the spotlight on scale, AI advisor Sagar Vishnoi, co-founder of Future Shift Labs, urges caution.

"Businesses and regulators should draw the line at tasks like scheduling and calendar work, which don't involve payments or entering other websites without human oversight or protocol. Many threats and harms have been identified recently, and they need to be identified, addressed and mitigated before any policy can be framed."

In other words, he believes agents should be limited, for now, to low-risk tasks like scheduling, and kept away from payments or logging into other sites without human oversight.

Speaking at NDTV World Summit, IT Minister Ashwini Vaishnaw reiterated the Indian government is working towards an AI policy and will float a consultation paper to invite suggestions in the coming weeks.

READ | AI Law In India Soon? What Ashwini Vaishnaw Said At NDTV World Summit

"Once a policy exists, organisations need their own policy for human oversight of agents. This should be framed for businesses first, and they should check these prerequisites before jumping onto the agents already offered by AI companies," Vishnoi told NDTV.

He also highlighted that the way Indians learn has changed. "Students, young people and small shop owners are learning from videos and reels and trying things hands-on. They will be deploying their own agents within this year or next."

However, the onus is upon people to keep their systems and financial information safe even without an AI policy in place, he said. "They should never give an agent their passwords, OTPs or other financial information."

Agents offered by companies right now need careful observation, and people should be cautious about using any agent without guidelines, an advisory or a policy, both nationally and in their own organisation, Vishnoi said.

The Degree Question

While Padmanabhuni and Vishnoi spotlight enterprises and safeguards, AI educator Ansh Mehra, founder of The Cutting Edge Group, took up the question at the heart of Walsh's teenager story: is a four-year degree still worth it?

"I agree that it should not take four years of a teenager's prime life to get a [B.Tech] degree, because the industry does not reward certifications. It rewards your ability to bring outcomes," Mehra told NDTV.

However, he cautioned that a degree is more than information. "AI is going to help any student learn faster, but it cannot replace the experience of being educated among a group of students. A four-year degree does not only promise you information, it promises an environment where a person can refine themselves and build the ability to navigate life," he said.

On Walsh's point that using AI in many courses can invite academic punishment, Mehra said institutions should stop treating AI as a separate subject and use it to understand existing ones. The bigger fix, he argued, is assessment.

"All teachers need to change the way they assess students. Currently, exams are like an event where students train to hack the system and get maximum marks. They optimise for marks, not for understanding the subject those marks are meant to reflect. Unless we change how we assess intelligence, people will keep chasing marks, and that is where they will use AI as a shortcut," he said.

As for the skills that matter, Mehra pointed to habits rather than tools. "Kids who read more, write more and articulate more will always be ahead in the AI age because, more than technical skills, your vocabulary defines how good you are with AI tools. The irony is that people are reading less, writing less and articulating less," he said.

"The future belongs to those who have good health, great taste, strong pattern-finding skills and the vocabulary to describe those patterns," he added.
Back on those two screens, a teenager's choice may be less binary than Walsh's picture suggests. Agents are coming, and the experts agree. The question is how fast and under what rules.