The Real Promise of Always-On AI Is Not Autonomy. It Is Negotiated Control.

OpenAI has introduced “dots”: AI agents designed to continue working in the background, connect across applications and return with results. The launch confirms a direction of travel, not an outcome. There is no public evidence yet that dots will raise productivity at scale, improve jobs or reduce them.

Still, the shift matters. A chatbot waits to be asked. An always-on agent can watch work, notice change and decide what deserves attention. That moves AI from being a tool people use towards becoming a participant in how organisations see, judge and organise people.

The decisive question is therefore not, “How intelligent is the AI?” It is, “Who sets its purpose, who can challenge its conclusions, and who receives the benefit?”

Always-on AI could relieve people of draining work. It might reconcile orders for a small business, watch for changes in regulation, help a disabled worker navigate inaccessible systems, or give an overstretched public servant more time with citizens.

But work is not merely a collection of tasks. Much professional judgment is built through apparently mundane activity: checking facts, drafting, resolving complaints and watching how experienced colleagues correct mistakes. If AI removes this layer too quickly, firms may save money today while weakening the apprenticeship system that produces tomorrow’s experts.

There is another risk. An agent connected to workplace conversations and documents can become an invisible manager. Its summaries may shape who appears productive, who seems slow and whose role looks unnecessary. A human may still make the final decision, but the AI may have framed the evidence.

People affected by these systems therefore need more than privacy promises. They need the right to know when AI has influenced a consequential judgment, to inspect the evidence, and to challenge errors before those errors affect employment, credit, healthcare or public benefits.

The International Labour Organization’s review finds that AI can save time, but those savings do not automatically become better output, higher earnings or better jobs. The result depends on how work is redesigned and how gains are shared.

An always-on agent may increase what one person or a small firm can accomplish. That could lower the cost of expertise and help smaller companies compete. For India, it could move technology and business-service firms from selling hours towards selling accountable outcomes. It could also give smaller enterprises access to analysis, compliance support and multilingual customer service that only large firms could once afford.

Yet productivity is not the same as prosperity. If workers produce more but wages stagnate, workloads rise and entry-level jobs disappear, the economy becomes more efficient without becoming more humane.

Power may also concentrate. Persistent AI depends on large providers that control models, computing capacity and access to organisational data. The World Bank’s World Development Report 2026 warns that developing countries must build the ability to adapt AI locally rather than merely import it. Otherwise, they may do the implementation work while much of the economic value flows elsewhere.

India’s opportunity is considerable, but so is the danger of a hollow middle: routine service work may shrink before enough higher-value roles and new learning pathways exist. The policy challenge is not to preserve every existing task. It is to ensure that the transition creates capability rather than dependence.

In the best future, AI removes drudgery, widens access to expertise and helps people make better decisions. Workers gain responsibility and share in the productivity dividend. Junior employees use AI as a tutor without losing the experience required to develop judgment.

In the most likely future, gains are uneven. Skilled workers and strong firms become more productive, while people with weaker bargaining power face closer monitoring, thinner career ladders and greater insecurity.

In the worst future, AI becomes an instrument of extraction: fewer people carry heavier workloads, automated judgments are difficult to contest, and most of the gains flow to owners and platform providers.

None of these futures is predetermined. Employers can preserve learning pathways and share gains. Governments can require transparency and avenues of appeal. Investors can reward businesses that create durable human capability rather than merely cutting headcount. India can invest in skills, local-language systems, public-interest capacity and the ability to audit the technologies on which it increasingly depends.