Retail & business banking
Answer like the branch. Prove it like the regulator.
A retail bank runs several high-volume queues that are almost entirely verification followed by a lookup and a status change. The conversation is the bottleneck, not the decision.
An illustrative reference architecture, not a customer
case study. No organisation is named and no deployment is described - this is
what the layers look like for a banking organisation,
so you can map it onto your own estate.
Channelswhere the conversation starts
Inbound numberspublished lines
SIP / contact centreexisting estate
Outboundcampaigns and callbacks
In-appbrowser WebRTC
▼
VoiceKernelthe agentic voice layer
Conversationagent, voice, turn taking
Knowledge groundinganswers with citations
Tool callsreads and writes back
Transferto a human, with context
Tenant isolationownership registry
Audit and erasureper call, per mutation
▼
Integrationyour boundary, your rules
REST + typed SDKscoped API keys
Signed webhooksretried, replayable
Live event streamSSE
▼
Bankingyour systems of record
Core bankingaccounts, balances, holds
Card managementblock, reissue, disputes
Fraud & AMLcase creation, risk signals
CRMinteraction history
IAM / KYCstep-up verification
Only the bottom layer is specific to banking. The
three above it are identical in every deployment, which is the point: the
voice layer does not need to know what industry it is in, only which systems
it is allowed to reach and what it may say.
What it changes
Mechanisms, not projections. Each of these follows from something structural
about the work, so you can check whether it holds for your own queues.
Verification stops being the bottleneckThe slow part of a card dispute is identifying the caller and the transaction, not deciding the outcome.
Peak-hour abandonment falls structurallyEvery call is answered at once, so service level stops competing with the staffing budget.
Regulatory evidence is a by-productFull transcript plus an audit row per mutation, produced without anyone remembering to write it down.
Staff move to hardship and complaintsThe conversations that need judgement get the people who have it.
What the agent actually does
Card disputesIdentify the transaction, capture the dispute reason, raise the case in the fraud system, block and reissue the card, and read back a reference number.
Balance and transactionsVerify, answer from the core, and offer a statement without a human ever joining.
Collections and hardshipExplain options, capture a commitment, and hand a warm transfer to a person the moment distress is detected.
What keeps it deployable
Every word auditedFull transcript and an audit row per mutation, with the actor and request id.
Verification before disclosureNo account data spoken until step-up succeeds; the tool call simply is not available before then.
Right to erasureRedaction in place, so the call stays auditable while personal data goes.
Card disputes, balance enquiries and payment status routinely make up the top three queues by volume in retail banking.
NextRead the source
Everything above is in the open-source Community edition. Nothing on
this page needs a licence.
NextRun it yourself
Point it at your provider account and put one queue through it before
you talk to anybody.
NextThen buy the guarantees
An Enterprise agreement adds the SLA and the
support contract when it carries real volume.