Product
02
AI Personas
Named agents that never improvise
Every Domu conversation is run by a persona: a named agent such as Taylor, Nubia or Sergio, with its own voice, tone, pacing and language. What the persona says is your script. AI classifies the customer's intent and selects the approved line for that moment in the call — it never writes one. Tone, pace and inflection flex; wording does not. Personas are assigned per segment, A/B tested against each other, and shaped by clients in Agent Studio through versioned approval. Domu runs in your name, on your approved scripts.
What it does
Ten ways a persona earns your trust
Named, differentiated personas
Distinct voice, pace, warmth and firmness per persona; matched to portfolio segment, product and geography, down to regional accent and local area code.
Live
Multilingual with mid-call switching
English, Spanish and Portuguese with Brazilian and Colombian cultural cues, switching language mid-conversation when the customer does.
Live
Non-bypassable guardrails
Verification-first flows, no balance disclosure before identity is confirmed, no rounding, hard-rule templates and stage-block logic shared by every persona.
Live
Red-flag detection
Lawsuit, threats, cease and desist, bankruptcy and hardship recognised in real time and routed to the right treatment.
Live
Warm transfer with context
Escalation to a human with a three-second hold that passes the AI summary to the agent, plus supervised mode with live barge-in.
Live · expanding
Persona A/B testing
Run personas head to head per segment. The winner is promoted and the result shown back in the Dashboard.
Live · expanding
Governed change
Versioning, maker-checker approval, rollback and a kill switch on every persona. New scripts and tactics are mined from your best collectors and added through that approval path — never learned live on a call.
Live
Agent Studio
Modify persona behaviour, scripts and boundaries from inside the platform, with automated evals before anything ships.
Coming soon
Classifier, not generator
AI classifies intent and picks from client-approved scripts. It picks the line. It never writes it.
Live
Bounded conversation stages
Greeting → verification → account or balance disclosure → payment or treatment options → wrap-up. Each stage carries a fixed set of intents and a fixed set of approved responses. Nothing can be said out of stage.
Live
How it compares
A guardrail engine the persona cannot talk around
Conversation model
Decision tree, or a generative assistant improvising inside a prompt
Classifier picks an approved line per stage; named personas carry the delivery
Compliance
Prompt instructions that drift over time
Guardrail engine outside the persona, rules cannot be talked around
Language
One language per flow with a hard handoff
EN, ES and PT with mid-call switching and local cultural cues
Hard cases
Escalate everything or miss it
Red-flag detection routes lawsuits, hardship and disputes correctly
Human handoff
Cold transfer, the customer repeats everything
Warm transfer with AI summary and live supervisor barge-in
Improvement loop
Quarterly script rewrites
Weekly A/B tests and automated evals; changes ship through maker-checker approval, not live learning
Change control
Vendor ticket
Versioned, approved, reversible, with a kill switch and Agent Studio
Proof
Tested like a product and measured like one
92%+
Conversations completed without a human
Collections BPO · proven
~90 days
Time to match and surpass the human benchmark
Collections BPO · proven
Under 1%
Calls with a severity-high QA flag
Target
+17%
Collections lift from a geography-matched persona
US bank pilot · proven
Customer result
Match and surpass the human benchmark in roughly 90 days
More products
The rest of the platform
Product
01
AI Dialer
Reaches the customer at machine scale, compliant at dial time
Learn more
Product
03
Intelligence Layer
Orchestrates your rules: who to contact, when, on which channel and with which persona
Learn more
Product
04
Omnichannel
Carries it across voice, text and email until it resolves
Learn more
Platform
Compliance
Certified before launch, guarded on every word and audited after every call.
See the trust layer
We’re building the next generation of engagement technology: intelligent, automated and compliant. Our mission is to empower financial institutions to orchestrate every stage of the servicing lifecycle with dignity and unprecedented efficiency.
Supported by
