Let's talk

The Agentic Era

How AI is Changing Society

As AI moves from answering questions to acting on our behalf, the fundamental structures of commerce, communication, and human roles are undergoing a massive shift.

By Divyaprakash JhaUpdated 6 min readTopic: AI and work

Photo for How AI is Changing Society

AI is shifting from a tool that answers questions to a system that takes action on our behalf. Over the next decade, this will transform how businesses operate, moving from simple chat interfaces to autonomous agents that manage complex, multi-day workflows.

The Rise of the Agent Economy

We are entering the "Agentic Era." Currently, most AI interactions are short-lived; you ask a question, and the AI gives an answer. However, the next phase of evolution involves AI agents that can hold a job. This means moving beyond simple drafts to "durable workflows"—tasks that take days or weeks to complete, such as following up with a sales lead, reconciling accounts every night, or preparing a complex tender.

By the 2027–2028 period, I forecast that these agents will become ordinary coworkers within business software. They won't just suggest text; they will manage processes that require memory, retries, and specific checkpoints where a human must approve the next step. This is a shift from "assistive" AI to "active" AI.

As we move toward 2030, the way businesses conduct commerce will change fundamentally. We expect a meaningful share of routine transactions—such as reordering stock, booking freight, or comparing supplier quotes—to be negotiated agent-to-agent. In this future, the "operating system" of a company will not be a collection of separate, disconnected apps. Instead, it will be a single AI layer that holds the company's data, rules, and agents to drive sales, inventory, and support.

Era Core Characteristic Primary Use Case
2024–2026 (Current) Assistive Chat Drafting emails, answering questions, summarizing text.
2027–2028 (Forecast) Durable Workflows Agents running multi-day processes with human sign-off.
2029–2031 (Forecast) Agentic Commerce Agents negotiating and transacting with other agents.

For builders, the goal is to design systems that can operate on a "ladder of trust." You start at the bottom, where the agent drafts and a human reads every word. As the system proves reliable through logs and error rates, you move up to "acting within limits," where the agent can perform tasks like issuing refunds under a certain amount without asking permission every time.

Language, Voice, and Global Access

The impact of AI will not be felt equally in all regions, and for many, the revolution will be spoken, not typed. In markets like India, the shift will be driven by voice-first technology. Many first-time internet users prefer speaking in local languages like Hindi, Marathi, or Tamil rather than typing in English. By 2028, I expect voice agents that answer calls, take orders, and explain bills in these local languages to become essential tools for small businesses.

This accessibility is being powered by two technical trends: cheaper inference and on-device models. Because the cost of running AI is dropping rapidly, we are seeing a move toward "small models." These capable models can run offline on a laptop or a mid-range smartphone. This is a game-changer for privacy and reliability, allowing people to handle everyday tasks without needing a constant, high-speed connection to a massive data center.

This technology has the potential to bridge massive gaps in social services. We are approaching a reality where a patient tutor is available in every child's native language, and a first-line health assistant can triage symptoms and explain medical reports in local clinics. For small businesses, this means automated customer service that is available at any hour, handling routine questions through voice and chat while escalating complex issues to the owner.

Sovereignty and National Infrastructure

As AI becomes central to economic productivity, it is no longer just a software concern; it is a matter of national security. Countries are increasingly treating compute power, data, and foundation models as strategic national assets. We are seeing a shift toward "Sovereign AI," where nations want to ensure their data stays within their borders and their industries aren't dependent on foreign-controlled infrastructure.

The IndiaAI Mission, approved in 2024, is a prime example of this trend. By funding shared GPU capacity and supporting the development of local foundation models, nations are attempting to build their own digital infrastructure. This ensures that the AI driving a country's economy reflects its specific linguistic and cultural nuances.

However, the physical reality of AI growth is hitting a wall. While we have spent years focusing on better algorithms, the next decade will be defined by physical constraints. By the early 2030s, the primary limit on AI expansion is expected to be the physical infrastructure: the availability of power plants, the capacity of electrical grids, and the global supply of specialized chips. Efficiency per watt will become just as important as the raw intelligence of the model.

What Remains Uniquely Human

A common fear is that AI will make humans obsolete. However, a more accurate view is that AI changes the *nature* of human value. As AI makes technical and repetitive tasks—like drafting code, data entry, or report assembly—significantly cheaper, the qualities that AI cannot replicate will become more valuable. These include judgment, trust, care, leadership, and the ability to decide what is actually worth doing in the first place.

AI can propose an outcome, but it cannot take responsibility. A human must still own the final decision and the consequences of that decision. In a world where anyone can generate a polished, professional-looking proposal in seconds, the competitive advantage shifts to "provenance." A trusted name, a real human relationship, and the ability to stand behind your work will be what customers pay a premium for.

This shift is already visible in how we think about work in the age of AI. We are moving away from a world where we are valued for performing individual steps in a process, and toward a world where we are valued for supervising the systems that execute those steps.

Human Skill Why It Matters in an AI World
Judgment Deciding if an AI's proposed outcome is actually the right goal.
Trust/Relationships Customers buy from people they believe will stand by the result.
Responsibility Taking the legal and moral blame when things go wrong.
Taste/Creativity Defining what is "good" or "meaningful" rather than just "correct."

Emerging Roles in an AI World

As old tasks are automated, entirely new professions will emerge to manage the new landscape. One of the most significant will be "Verification." Just as we have financial auditors to ensure companies follow accounting rules, we will need AI auditors. These professionals will test agents, review execution logs, and sign off that a system is behaving exactly as it was claimed to behave.

We will also see a surge in demand for roles that sit at the intersection of human intent and machine execution. This includes:

  • Workflow Designers: People who break down complex business processes into the steps, inputs, and approval gates that an AI can follow.
  • AI Operations Managers: Experts who manage a fleet of agents, handling the exceptions and "edge cases" that the AI cannot resolve.
  • Compliance & Ethics Officers: Ensuring that autonomous agents operate within legal and ethical boundaries, especially when they have the authority to spend money.

The nature of work is shifting from "doing" to "directing." Instead of spending eight hours a day typing invoices or checking leads, the worker of the future will spend their time designing the system that does those things and verifying that it did them correctly.

Societal Risks to Manage

The transition to an agentic society is not without significant risks. One of the most pressing technical threats is "prompt injection." This occurs when a hidden instruction—perhaps buried in a website or a document—hijacks an AI agent. If that agent has the authority to send money or change customer records, a single malicious prompt could cause significant financial or operational damage. This makes "least privilege" security (giving agents only the minimum access they need) a critical requirement.

On a broader social level, we face the challenge of misinformation. As it becomes incredibly cheap to generate realistic fake text, voice, and video, the risk of fraud increases. We must develop better ways to verify the provenance of content so that people can distinguish between a real person and a sophisticated digital clone.

Finally, there is the risk of "skill atrophy." If businesses and individuals hand every critical task over to AI, we may lose the foundational human skills required to perform those tasks manually. This becomes a major vulnerability if a vendor changes their pricing, a model is updated, or a system goes offline. To stay resilient, we must keep humans practiced in the critical work that keeps the world running.

Questions people ask

Will AI replace humans in society?

AI is moving from answering questions to acting on our behalf. While it will automate repetitive tasks, it will likely increase the productivity of roles that require judgment, relationships, and responsibility.

How will AI change communication?

Communication will become increasingly voice-first and localized, especially in markets like India, where users prefer speaking in native languages over typing in English.

What are the main constraints on AI growth?

By the early 2030s, the primary limits on AI expansion are expected to be energy availability, power grid connections, and the supply of chips.

What is an AI agent?

An AI agent is a system that goes beyond simple chat to perform durable workflows, such as managing a multi-day process with memory, retries, and checkpoints.

How can I prepare for an AI-driven society?

Focus on developing 'compounding skills' like clear writing, deep domain expertise to verify AI output, process design, and interpersonal skills like negotiation and care.