Punjab & Sind Bank — Punjab & Sind Bank Takes Enterprise GenAI from PoC to Production

Executive summary

Punjab & Sind Bank uses SimpleWorks’ R-YaBot as a production RAG assistant for employee access to the bank’s approved knowledge corpus. Employees use it for quick lookups and for longer research tasks such as reconciling a circular, checking a policy chain and preparing a customer response.

Over eight months, the deployment achieved nearly 95% adoption across more than 4,000 licensed users and handled more than 75,000 questions in production. Punjab & Sind Bank also received the Best Use of AI in Banking & Financial Services award at the Bharat NBFC & FinTech Summit & Awards 2026.

Bank employees need to work with circulars, policies and other approved documents while carrying out both quick checks and more involved tasks. In regulated banking, an AI answer cannot merely sound reasonable. It must be grounded, verifiable and trustworthy.

The deployment therefore had to support everyday employee work without presenting unsupported material as fact, while also giving users a clear path back to the bank’s source documents.

What employees use R-YaBot for

• Single lookups: Employees use R-YaBot to retrieve an individual fact or document-based answer.
• Circular research: Employees use it to work through circulars rather than retrieving only one fact.
• Policy-chain checks: Employees use multi-question sessions to follow requirements across related policies and documents.
• Customer-response preparation: The review identifies preparing a customer response as one of the tasks supported by longer sessions.
• Source verification: Employees can open the underlying circular or policy document to check the answer before relying on it.

R-YaBot therefore supports both quick lookups and longer tasks, including reconciling a circular, preparing a customer response and checking a policy chain.

What employees actually ask

Employees type into R-YaBot the way they type into a search box. The production log is mostly two or three word entries such as "eblr" or "gold loan", not full sentences. The examples below show some of the most frequently raised queries alongside the question the employee was actually asking.

• “loan policy” → Where can I find the latest loan policy?
• “staff housing loan” → What is the staff housing loan scheme?
• “personal loan” → What are the details of the personal loan scheme?
• “gold loan” → What are the terms of the gold loan scheme?
• “gst ease” → What is the GST Ease facility?
• “legal audit” → What is the legal audit process?

What stands out in this pattern of use is breadth. Employees ask about loan products and specialised agri and MSME schemes, circulars and policy lookups, interest rates, branded retail schemes, staff benefits and allowances, accounts and deposits, compliance and audit processes, portal and system links, and service charges. They are not using R-YaBot for one narrow task. They are using it across the bank's operating knowledge.

Because every query is logged, the pattern of demand is itself an output. Topics that employees ask about repeatedly confirm which parts of the approved corpus are doing the most work, and questions that go unmatched show the bank where additional approved content is needed.

The SimpleWorks solution

R-YaBot operates as a RAG assistant using Punjab & Sind Bank’s approved knowledge corpus, which is curated and version-controlled by bank teams.

• Approved knowledge corpus: R-YaBot operates using the bank’s approved knowledge corpus.
• Source traceability: Users can trace answers back to the relevant circulars, policies and documents.
• Answer verification: Employees can open the underlying source documents to verify the answers.
• Unsupported answers: If an answer is not supported by the approved knowledge base, R-YaBot states this rather than coming up with one.

Value delivered

• Approved information in the flow of work: Employees can use the same assistant for a quick lookup or a longer document-based task.
• Verifiable answers: More than 80% of questions were answered directly from approved bank documents. Users opened source documents more than 5,500 times to verify answers.
• Safer handling of unsupported questions: Questions outside the approved knowledge corpus return “not in my documents” rather than an unsupported answer.
• A demand-led content roadmap: Unmatched questions reveal where employees need additional approved content and can be grouped by theme to guide future corpus expansion.
• Sustained production adoption: The deployment achieved nearly 95% adoption across more than 4,000 licensed users and handled more than 75,000 production questions.

These results demonstrate access, adoption, groundedness and verification. The supplied evidence does not quantify time savings, cost reduction, revenue impact or customer-satisfaction improvement, so no such claims are made here.

Results and adoption

Over eight months, R-YaBot achieved nearly 95% adoption across more than 4,000 licensed users and handled more than 75,000 questions in production.

More than 80% of those questions were answered directly from approved bank documents. Employees opened the underlying source documents more than 5,500 times to verify answers, while questions outside the approved corpus returned “not in my documents” rather than an unsupported answer.

Industry recognition

Punjab & Sind Bank received the Best Use of AI in Banking & Financial Services award at the Bharat NBFC & FinTech Summit & Awards 2026.

For SimpleWorks, the award reflects the importance of taking GenAI beyond proofs of concept and incorporating it into secure, governed and everyday production use.

Building trustworthy AI for regulated enterprises

This deployment reflects SimpleWorks’ approach to AI: grounded, verifiable systems that achieve sustained production adoption in regulated enterprises. SimpleWorks is proud to act as Punjab & Sind Bank’s technology partner on this journey.