"Allowing AI agents in consumer grievance redressal while mandating human oversight reflects a calibrated regulatory approach." Examine with reference to TRAI's 2026 amendment.

TRAI's Fourth Amendment (November 2026) to the Telecom Consumers Complaint Redressal Regulations, 2012 makes Complaint Centres run 24x7. It allows AI agents, provided there is "appropriate human oversight" and humans are available for at least 16 hours a day [1]. The design is calibrated in intent, but in practice it has gaps.

Why the approach is calibrated

• Access and efficiency together: AI covers the night-time gap for shift workers and rural users. The 16-hour human minimum protects consumers who are less comfortable with digital tools [1].
• Human-in-the-loop: AI is allowed but not compulsory, and it is never left without supervision. This is a risk-proportionate middle path between banning AI and leaving it unregulated [1].
• Inclusion safeguards: PwD desks or helplines are mandatory. The app, website and portal must meet GoI accessibility standards [1].
• Outcome checks: these include a five-point Feedback Survey published quarterly, reports to the CEO or Board, an Oversight Committee with consumer members, a TRAI audit, and financial disincentives [1].
• Consultative legitimacy: the draft (May 2026) noted that consumers were moving to apps and chatbots. Comments and an Open House Discussion followed [2][1].

Where the calibration falls short

• Undefined oversight: the term "appropriate" is not defined. For up to 8 hours a day, each provider decides how closely humans review the AI's replies [1].
• Self-grading: the provider runs the survey and files its own Quarterly Performance Report (QPR). So the company that deploys the AI also grades its performance [1].
• Closed is not the same as solved: on CPGRAMS, only 69.8% of about 6 lakh resolved grievances were rated satisfactory [4]. An AI tuned for speed could close cases without fixing the problem.
• No neutral appeal: the Appellate Authority is staffed by the provider's own managers [1]. TRAI has no mandate to settle individual complaints [3][5], so a consumer harmed by an AI decision has no independent referee.

Way forward

• Define the minimum level of oversight: consumers should have a right to reach a human on request, and AI decisions should leave an audit trail.
• TRAI should run random third-party checks of survey and QPR data, using its existing audit powers [1].
• Create a statutory, independent third tier above the providers' two tiers.

Overall, the amendment's principle is sound: AI widens access while humans stay accountable. Clear oversight standards, independent checks of the data and a neutral appeal forum would turn this into real calibration. Together they would also fulfil the Citizen's Charter promise of redressal that is both fast and fair.