When code makes decisions

India is ready to deploy AI agents faster than it is ready to govern them. In boardrooms, AI conversations are still dominated by use cases and productivity gains. Far less attention is paid to a harder question. What authority will these systems hold?

AI agents differ from conventional assistants. They do not merely generate content. They can retrieve information, interpret it, recommend actions, communicate with stakeholders, and trigger workflows across enterprise systems. The risk does not begin when a machine starts running a company. It begins earlier, when decision-making authority is handed over one permission at a time.

For example, a finance agent may first identify mismatched invoices. It may later contact vendors, recommend adjustments, and route transactions for approval. A customer-service agent may move from drafting replies to issuing refunds within limits. The progression appears incremental, but the shift matters. Software is no longer only supporting a process. It is exercising authority inside it.

That shift could happen quickly in India. Enterprises are digitising, cloud usage is expanding, and the country has deep technology talent and a software ecosystem.

More importantly, India has repeatedly shown that it can compress technology costs while expanding value at scale. Mobile data became available at a fraction of global price levels, while digital payments moved towards near-zero marginal transaction costs.

These are strong foundations for adoption. However, they do not, by themselves, amount to readiness. Many business processes remain fragmented, poorly documented, or dependent on tacit human judgment.

Autonomous AI is therefore as much an operating-model challenge as a technology challenge. This matters particularly when AI enters finance, compliance, or reporting. Once an agent begins reconciling accounts, resolving exceptions, or preparing regulatory submissions, it is no longer sitting alongside the control environment. It has become part of it.

Existing controls usually assume that a human initiates, reviews, or approves an action. AI agents break that assumption. Companies must define which data an agent may access, which systems it may enter, which actions it may take, and the limits within which it can operate.

Readiness will vary sharply across India Inc. A large bank may have model risk, cyber, audit, and regulatory controls to build upon. A mid-sized enterprise adopting agents via SaaS platforms may lack that scaffolding. However, accountability must remain with the enterprise.

A company cannot transfer responsibility for a business outcome simply because the model or software came from a third-party vendor. Regulation may define minimum obligations, but it cannot design an enterprise’s approval limits, escalation logic, or accountability model.

A practical readiness test comes down to five questions. Where are agents deployed? What authority do they hold? Who owns each material outcome? Are important actions recorded? Can a human override or stop the system?

India will adopt enterprise AI agents quickly. If companies combine the country’s instinct for affordability and scale with equal discipline in process design, governance, and accountability, India may do more than participate in the AI revolution. It may redefine the economics of AI adoption worldwide.

The writer is co-founder & CEO, Uniqus Consultech

Disclaimer: The views expressed are the author’s own and do not reflect the official policy or position of Financial Express.