In-App AI Agents for Stock Broking Apps in India: Onboarding, Activation, and SEBI Rules

A stock broking app can use an in-app AI agent to get more new customers from an opened account to a funded account and a first trade: it sees where each customer has stalled, explains the next step in plain language, and takes them to the right screen inside the app. Before deploying one, settle three things with compliance: the agent explains features and processes but never recommends a security, a trade, or a segment; it never shortens or works around SEBI's mandated risk disclosures; and the broker owns everything the agent says, because SEBI's Intermediaries Regulations make a regulated entity solely responsible for the output of the AI tools it uses, including tools bought from a vendor.

Where New Broking Customers Stall Before Their First Trade

Most of the loss happens after sign-up and before the first order, at steps where the customer has to do something they do not fully understand.

• KYC status: the customer's KYC record is pending, on hold, or needs modification, and the app shows a status label without saying what to do next.
• Account opening steps: document upload, e-sign, bank account verification, or in-person verification fails or is left half finished.
• Segment activation: the customer wants derivatives or another segment, and the extra documents and declarations are unclear.
• First funding: the account is open but has no money in it, often because the customer is unsure how funds, margin, and settlement work.
• First order: the customer reaches the order screen and stops at order types, product types, or charges they have not seen before.

What an In-App Agent Does at Each Step

The agent is useful because it can see the customer's state inside the app, not only what they type.

• It reads the journey state (KYC status, which opening step failed, whether the account is funded) and opens with the specific next step, not a generic greeting.
• It explains a rejected document or a failed verification in plain words and takes the customer back to the exact screen to fix it.
• It answers process questions about funding, settlement, charges, and order types from the broker's own approved content, and shows the relevant screen.
• It hands over to a person, with the conversation attached, when the customer has a complaint, a dispute, or a question the approved content does not cover.

The app stays the place where the customer acts. The agent works inside it and makes each screen easier to finish.

What the Agent Should Not Do

A broking app sits next to activities that are regulated in their own right, so the boundary matters more than in most financial apps.

• No recommendations: the agent does not suggest which stock, fund, contract, or segment to choose, and does not comment on whether a trade is a good idea. Agree the exact boundary between explaining a feature and giving advice with compliance, and test it with real customer questions.
• No nudging into derivatives: the agent can explain how segment activation works when the customer asks, but it does not prompt customers toward derivatives trading.
• No shortcuts on consent: the customer completes KYC, e-sign, and declarations themselves. The agent guides, it does not act on the customer's behalf.

What SEBI's Rules Mean for an In-App Agent in a Broking App

Three SEBI documents shape the design. Check each against its current version with your compliance team before launch.

• Responsibility for AI output: the SEBI (Intermediaries) (Amendment) Regulations, 2025, notified on 10 February 2025, insert regulation 16C. It makes any person regulated by SEBI that uses AI and machine learning tools, whether built in-house or procured from a third party, solely responsible for the privacy, security, and integrity of investor data, for the output of those tools, and for compliance with applicable laws. A vendor contract does not move that responsibility.
• Risk disclosures for derivatives: SEBI's circular of 19 May 2023 (SEBI/HO/MIRSD/MIRSD-PoD-1/P/CIR/2023/73) requires stock brokers to show clients the prescribed risk disclosures on equity futures and options, which may appear as a pop-up on login, covering at least 50% of the screen, with the client allowed to proceed only after acknowledging them. An in-app agent must never cover, summarise away, or click through that disclosure, and should not appear on top of it.
• Proposed AI/ML guidelines: SEBI's consultation paper of 20 June 2025 on responsible usage of AI/ML in securities markets proposes, among other things, that market participants disclose the use of AI/ML to clients where it directly affects them, including in advisory and support services, in language clients can understand. It is a consultation paper, not a final rule, so check whether SEBI has since issued final guidelines.

What Your Team Needs to Provide

An in-app agent in a broking app is only as accurate as the data and content it is given.

• Journey events and status: KYC and account opening status, the step that failed and the reason code, segment status, and funding status, read at the moment the customer opens the agent.
• Approved content: answers on charges, settlement, funding, order types, and segment activation, owned by product and approved by compliance, with a version history.
• Deep links into the app: the agent should take the customer to the exact screen for the next step rather than describe where it is.
• A blocked-topics list and logging: questions that touch on recommendations route to a fixed response or a person, and every conversation is logged so compliance can review it.

How to Run a Pilot

Pick one stalled cohort, for example customers whose account opened in the last 14 days but who have not funded it, and keep a randomly selected holdout group on the current app experience. Choose one primary metric, such as the share of the cohort that funds the account within 30 days. Report the rate among customers who used the agent separately from the difference between the agent group and the holdout: the first is an observed rate, and only the second shows what the agent added. Review a sample of transcripts every week for anything that reads like a recommendation, and treat a single confirmed case as a stop-and-fix issue, not a metric.

When an In-App Agent Is the Wrong Fix

If customers stall because a verification partner is down, a step is broken on certain devices, or the charges themselves are the objection, an agent will explain the problem more politely but will not remove it. Fix the journey first. An agent is also premature if you cannot yet log and retrieve every conversation by client, or if compliance has not agreed the advice boundary in writing.

Where RevRag AI Fits

RevRag AI's In-App AI Agent is embedded inside a financial app, reads where the customer is in a journey, and guides them through the next step on the same screen. For broking and wealth platforms, that means account opening, activation, and first-funding journeys inside the app the customer already uses. If you are planning a pilot, you can book a demo and bring the stalled cohort you want to start with.

Frequently Asked Questions About In-App AI Agents for Stock Broking Apps

Should the agent tell customers that it is an AI agent?

Yes, plainly and at the start. Regulation 16C does not prescribe wording, but SEBI's June 2025 consultation paper proposes that AI use which directly affects clients be disclosed in language they can understand, and a clear label avoids any suggestion that a person, or an adviser, is answering.

Can the agent answer questions about derivatives?

It can explain how segment activation works and what the app's screens mean, using approved content. It should not suggest that a customer trade derivatives, and it must leave the mandated risk disclosures exactly as SEBI and your exchange require them to be shown.

Who is responsible if the agent gives a wrong answer?

The broker. Under regulation 16C the regulated entity is solely responsible for the output of AI tools it uses, including tools procured from third parties. Put content approval, transcript review, and a fast correction process in place before launch, and cover the vendor's obligations in the contract.

How is this different from an AI agent in a mutual fund app?

The journeys differ. A mutual fund app agent mostly works on risk profiling, SIP setup, and mandate registration. A broking app agent works on account opening, segment activation, first funding, and the first order, where the boundary between explaining and advising is closer and the derivatives disclosures apply.