Bank of Baroda’s AI rollout

Bank of Baroda has rolled out a new artificial intelligence system that lets a customer in one Indian language speak to a bank employee in another, live at the branch counter. The bank says the platform, called bob SAMVAD, handles real-time two-way conversations across 22 Indian languages and was built in-house rather than bought from an outside vendor. (pib.gov.in) That sounds simple until you picture the actual branch. A customer asks about a pension credit in Marathi, a staff member is more comfortable in Hindi or English, and the conversation has to stay accurate enough to move money, update records, or explain a product without mistakes. (pib.gov.in) Banking in India has always had this language problem hiding in plain sight. India recognizes 22 scheduled languages at the national level, and a large public-sector bank can serve customers across thousands of branches spread across states where the local language changes every few hundred miles. (pib.gov.in) Bank of Baroda is exactly the kind of institution where that friction shows up every day. As of March 2025, the bank operated more than 8,500 domestic branches, along with more than 9,316 automated teller machines and over 21,000 business correspondent outlets, which means a language gap at the counter is not a niche problem but a mass-scale operations problem. (economictimes.indiatimes.com) The first phase of the rollout is not nationwide yet. Multiple reports say bob SAMVAD is being introduced first in 250 branches across five southern and western states, which suggests the bank is testing the system where language diversity and branch traffic can stress the model before a broader expansion. (techobserver.in, ibef.org) The mechanics are straightforward on paper. Customers can speak or type in their preferred language, the system translates the request into the staff member’s chosen language, and the reply can come back as text or optional voice output for the customer. (pib.gov.in, economictimes.indiatimes.com) The harder part is not translation but banking accuracy. A missed word in a movie subtitle is annoying, but a missed word in a nominee update, a loan explanation, or a complaint about a failed transfer can create a compliance issue, a reconciliation issue, or a customer dispute that has to be cleaned up later. (rbi.org.in, pib.gov.in) That is why this launch fits a larger shift in banking technology. Banks are no longer using artificial intelligence only for back-end scoring or fraud flags; they are pushing it into the customer conversation itself, where the system touches disclosures, consent, service quality, and recordkeeping in real time. (rbi.org.in, economictimes.indiatimes.com) India’s central bank has been preparing for exactly this kind of shift. In August 2025, the Reserve Bank of India published its Framework for Responsible and Ethical Enablement of Artificial Intelligence report, which discusses governance, operational risk, resilience, incident reporting, and oversight for artificial intelligence used in financial services. (rbi.org.in) That framework matters because a multilingual branch assistant is not just a nicer interface. It becomes part of the bank’s control environment, which means managers need to know what was said, how it was translated, whether the response matched policy, and how exceptions are reviewed when the machine gets something wrong. (rbi.org.in) There is also a quiet staffing angle here. A system like bob SAMVAD does not replace the branch employee so much as it changes the employee’s job from “speak every customer’s language” to “verify the transaction, explain the rule, and catch the edge case,” which pushes more value into supervision and less into rote translation. (pib.gov.in, economictimes.indiatimes.com) For customers, the appeal is immediate. Elderly users, rural customers, and first-time formal banking users often need branch help for tasks that are too important or too confusing to leave to an application menu, and being able to ask those questions in a familiar language can reduce hesitation at the exact moment trust is being built. (pib.gov.in, thehindubusinessline.com) For the bank, the bet is that better conversations produce cleaner operations. If the translation is fast, accurate, and logged properly, the branch can serve more people without hiring for every language combination, but if the controls are weak, the same tool can multiply small errors across hundreds of branches just as quickly. (rbi.org.in, techobserver.in) So the real story is not that Bank of Baroda added an artificial intelligence feature on April 7, 2026. It is that one of India’s biggest public-sector banks has started treating live customer conversation itself as software, and once that happens, language access, compliance controls, reconciliation trails, and branch operations all become part of the same system. (pib.gov.in, rbi.org.in)