Flominzo

Flominzo Voice

Banking with voice and agent

Send, track and reconcile money by talking. The agent understands; the core decides.

Built on Flominzo, the agentic payment OS live with remittance companies. Voice is in pilot.

Last updated:

Flominzo VoiceIllustrationSpeaking
You:
Send 5000 to Ramesh again.
Flominzo agent:
₹5,000 to Ramesh Kumar at HDFC Bank, ending 4821. Fee AED 3.00. Please confirm in your app.

Confirm in your app

To
Ramesh Kumar · HDFC ••••4821
Ramesh gets
₹5,000.00
Fee
AED 3.00
You pay
AED 216.22
Confirm with biometrics
An illustration with sample data. Money moves only after the confirmation in the app.

Who it is for

For people who would rather talk.

One core for senders and ops: the same evidence store answers “where is my money?” and “why did this payout break?”

  • Senders

    Speak instead of typing

    Migrant workers sending money home, often in mixed languages and on low-end phones. They ask where their money is, and send again to people they already pay.

  • Ops teams

    Ask why, get the evidence

    Operations teams working reconciliation breaks, delayed payouts and partner failures. They ask why a payout broke and get the mismatch and the evidence event.

How a voice request moves money

Voice is a channel, not a credential.

Voice never authenticates on its own: every voice request becomes a PaymentIntent that passes the same checks as app and API. Read the security principles.

  1. You speakIn plain English today. Other languages are integrated per deployment.
  2. The agent understandsIt resolves the words to one intent: who, how much, what for.
  3. A PaymentIntent is createdIt passes the same deterministic checks as the app or the API.
  4. Checks pass?Yes: continue.No: Declined. The reason is logged.
  5. You confirm in the appThe agent reads back amount, beneficiary and fees. OTP or biometric on your registered device.
  6. The core executesThe payout goes to the partner, with every step recorded.

Event store: every step, including declines.

A voice request becomes a PaymentIntent, passes the same deterministic checks as app or API, and only moves money after the user confirms on their registered device; every step, including declines, lands in the event store.

Use cases

Five questions, answered from evidence.

Three for senders, two for operations. Each card shows what is asked and what the answer contains.

  • Sender

    Transfer status

    “Where is my money to Mum?”

    The answer: Its current state, the expected arrival, and the partner it is with.

  • Sender

    Send again to a saved beneficiary

    “Send 5000 to Ramesh again”

    The answer: A read-back of amount, beneficiary and fees, then “confirm in your app”.

  • Sender

    Check a beneficiary before sending

    “Which account do I have for Ramesh?”

    The answer: A masked account, the bank name and the last transfer date.

  • Operations

    Reconciliation break

    “Why did payout 4471 break?”

    The answer: Operational and reconciliation state, the mismatch found, and the evidence event.

  • Operations

    A delayed or failed partner batch

    “What is stuck with the partner from yesterday?”

    The answer: The count, the partner and the settlement-cycle reason, such as a 23:59 payment settling at 00:01.

Languages

English today. Any language by integration.

Flominzo Voice runs in English today. It is built to work in any language: speech recognition, the agent’s understanding and the spoken read-back are integrated for the languages your senders use.

Tell us the languages your senders speak, and we scope them for your deployment.

Integrated per language

  • Speech recognition for the language and accent
  • The agent’s understanding of requests, including two languages mixed in one sentence
  • The spoken read-back and replies

The same in every language

  • The PaymentIntent and the deterministic checks
  • Confirmation in the app, on the registered device
  • The event store and the evidence behind every answer

Security and trust

The money never moves on a voice command alone.

Voice is a channel, not a credential. Every request follows the flow above.

  • Can a spoken yes confirm a payment?

    Voice never authenticates on its own.

    Login stays device-bound; confirmation is OTP or biometric in-app, never a spoken yes.

  • Can the agent move money?

    Agents investigate and explain.

    They never move money.

  • Is a voiceprint used to log in?

    Voiceprints, if ever used, are a risk signal only, never a key.

  • Will the agent read out account numbers or OTPs?

    Sensitive data is never spoken aloud:

    no full account numbers, no OTPs, no balances on speaker unless the user asks and the device is verified.

What is logged

  • The utterance transcript, if you consent
  • The detected language
  • The resolved intent
  • Every check result
  • The read-back
  • The confirmation event
  • The payout partner’s response

What is never spoken aloud

  • Full account numbers
  • OTPs
  • Balances on speaker, unless you ask and the device is verified

Built on Flominzo

One core under every channel.

Voice is one more way into the Flominzo core. Three layers do the work: the agentic app turns speech into a PaymentIntent, the payout partner integration carries the money, and reconciliation proves what happened.

For engineers, the voice layer is an API surface on the same PaymentIntent core as app, web and WhatsApp.

Put voice on one corridor.
The core still decides.

Start with one corridor in English, and add the languages your senders speak. Every voice request passes the checks you run today, and money moves only after the confirmation in the app.