New Delhi: The most important interface of the last decade, a smartphone app, might be nearing its end.
“There are probably 15 million applications on the app store (Apple and Android stores)…In the next five to ten years, they could all vanish,” said Nikesh Arora, chief executive officer (CEO) of cybersecurity company Palo Alto Networks, at a side event last week during the AI Impact Summit in Delhi.
He believes they will be replaced by AI agents that handle tasks using natural language.
Arora said today’s apps are essentially rigid workflows designed for humans with forms to fill, buttons to click, steps to repeat. Booking a flight requires entering names, dates and destinations. Ordering a cab means switching apps, and making a restaurant reservation requires yet another app.
He argued that with AI agents, this structure could change to conversation-based workflows. Instead of flipping between airline, hotel, taxi and restaurant apps, users could command a single “super agent.”
This agent would manage the whole process: booking the ticket, reserving the room, scheduling the ride, and getting dinner. It would fill in missing details, ask for clarification and connect directly to backend systems.
“What is most convenient for a human being is to tell your phone to do these four things,” Arora said. In that world, multiple agents and not multiple apps complete the task.
The implications of this shift are significant for India, where app-based startups lead sectors ranging from ride-hailing and food delivery to fintech and e-commerce. If user interfaces shift from apps to conversational agents, companies may be reduced to backend service providers accessed through a third-party AI layer.
However, this future possibility raises new regulatory and competitive questions.
Melanie Dawes, chief executive of UK regulator Ofcom, who shared the stage with Arora, challenged the notion that a universal agent layer would simply democratize access.
“The super agent will then take a cut,” she said, highlighting concerns around concentration of power and consent.
Dawes asked: if an AI agent books a hotel incorrectly or fails to dispatch a taxi, who is liable—the service provider, the model developer, or the agent orchestrator?
For nearly two decades, companies have competed to keep users inside their proprietary apps. The data generated and used by these apps is also used for targeted ads and personalised user experience.
If conversational AI becomes the dominant interface, the battle is likely to move up a layer toward controlling the agent that mediates access to every service.
