Conversational payments: moving money in natural language

The simplest promise of conversational banking is also the most demanding: "Send 10k to my landlord" should just work. No forms, no account-number typing, no menu diving. Conversational payments turn a sentence into a completed, receipted transfer, and doing that safely is real infrastructure.

What a payment sentence contains

"Send 10k to my landlord for August rent" packs in everything a transfer needs:

• Amount: 10,000
• Recipient: "my landlord", resolved to a saved beneficiary
• Purpose: August rent (the narration)

The system extracts these with natural-language understanding, resolves the recipient, and prepares the transfer, then asks the customer to confirm before anything moves.

The recipient problem

The riskiest part of any payment is who receives it. Conversational payments have to resolve names to accounts carefully: match saved beneficiaries, handle "the same person as last time," and never silently pick the wrong "Ada." When there is doubt, the assistant asks. The confirmation step is the customer's last, clear checkpoint.

Beyond one-off transfers

Once payments are conversational, more becomes natural:

• Repeat sends. "Send him the same as last week."
• Split actions. "Pay my rent and send 5k to my sister."
• Scheduling. "Send this every month." See automating recurring payments.

Each is just a sentence, backed by the same tool-calling and guardrails.

Why it is better than a form

Forms make every payment the same amount of work, whether you are an expert or a first-timer. Conversation collapses that: the more the system knows you, the shorter your instructions get. That is the opposite of a menu, which gets more crowded as you add features.

Conversational payments are where most people first feel the shift. TwelveAI provides the infrastructure underneath. See the features.