Vaada (वादा) means promise
Companies make promises on the phone. Vaada keeps them.
हर वादा, लिखित में।
Indians sort out failed payments, wrong bills, and missing refunds by calling a helpline, usually in Hindi, English, or both at once. The call ends with a promise and leaves no record. Vaada listens to the call, writes down every promise with the exact words it was made in, checks it against what the regulator already requires, and drafts the follow-up and the escalation in the consumer's own language.
One failed UPI payment, start to finish
A fictional call, replayed deterministically: no model request, no microphone, nothing stored. Every claim on the right links back to an exact line on the left.
Transcript
Try it on your own failed UPI payment
Enter the transaction and paste anything the company told you, in English or Hinglish. Vaada finds the promises, applies the RBI deadline as of today, and drafts the written complaint. It runs entirely in your browser; nothing is sent or stored.
Your result
Your deadline, compensation, promises and draft will appear here.
Where it fits in Bessemer's India portfolio
The same engine, evidence-backed promises with a human decision before anything goes out, works on both sides of the support call. These are hypotheses, not claims about any company's data.
Claims are a chain of promises
Device-protection and warranty claims, the business Zopper runs for retailers and OEMs, move through pickup, assessment, and settlement dates promised on calls. Run inside a claims team, Vaada would show which promises are about to slip before the customer has to chase them.
Refunds and re-service at scale
For high-volume consumer platforms such as Swiggy, BigBasket, and Urban Company, a promised refund or a free re-service visit is a small commitment made thousands of times a day. Tracking each one against its source quote turns support quality into something a team can measure.
Close the loop with consent
Today Vaada trusts the consumer to say whether the money arrived. On India's consent-based Account Aggregator rails, the kind of financial-data infrastructure Perfios builds, a user could let Vaada confirm the refund actually landed and stop the escalation automatically.
Why this, why India, why now
The thesis in three parts. Each is a reason the product could work here before it works anywhere else.
Support happens by voice, in mixed languages
For most Indian consumers the real interface to a bank, telecom, or insurer is a phone call or a WhatsApp voice note, and the conversation switches between Hindi and English mid-sentence. Text-first complaint forms miss that user. A voice copilot that understands Hinglish meets them where the problem actually gets discussed.
UPI made payment disputes a mass-market problem
When everyday rent, groceries, and transfers run on UPI, a failed transaction is no longer a rare banking edge case. It is something ordinary households hit, and a few thousand rupees stuck for a week is real money and a real late fee.
The rules already exist. Awareness doesn't.
The RBI has already set reversal deadlines and automatic compensation for failed transactions, and an Ombudsman route when a bank does not respond. Most people never use them because they do not know the dates or the wording. Vaada turns existing regulation into a consumer product instead of waiting for new regulation.
What I would test first
Three experiments, cheapest first. Each has a number that decides whether to keep going.
| Experiment | How | Keep going if |
|---|---|---|
| Do people want the record? | A WhatsApp number that turns a forwarded voice note about a support call into a dated list of promises and the RBI deadline, for failed UPI payments only. | A meaningful share of users come back with a second dispute within 60 days, without paid acquisition. |
| Does the letter change the outcome? | Compare time-to-refund for users who send the Vaada-drafted written complaint against users who only call again. | Written, rule-citing complaints resolve faster or more often, so the product is doing something a phone call can't. |
| Who pays? | Test willingness to pay for escalation help on larger disputes (insurance claims, telecom billing) and test the same engine with one bank or fintech that wants its own agents to keep promises. | Either consumers pay for high-stakes cases, or a company pays to use promise tracking on its own support floor. |
How it's built
Vaada is a consumer mode on top of Conversation Copilot, the open repository behind this page. It keeps the engine's rules: evidence before automation, a human decision before anything is sent, and nothing saved without consent.
Listen and translate
The Live copilot already captures audio, transcribes it, and translates mixed-language speech in real time. Vaada uses that path for Hinglish calls.
Extract with evidence
The Close the Loop engine turns a transcript into commitments that each carry an exact source quote. Vaada adds regulator timelines on top and checks every quote against the transcript in tests.
Learn from decisions
Approvals, edits, and rejections become evaluation data in the Eval Control Tower, so the drafts get better at the cases people actually escalate.