Educate and answer
It explains how things work and answers questions grounded in your own knowledge base, not the model's general training. If nothing relevant is found, it hands off rather than inventing an answer.
Client Concierge - the inbound AI agent
Built - web chat liveA 24/7 branded worker on the public edge of your funnel.
The Client Concierge answers a client's questions, runs real calculations, looks up indicative loan products, qualifies the enquiry, writes a real enquiry, and takes a booking request, then hands a serious client to a broker with full context. Every touchpoint is under your brand, and every figure it quotes comes from a real calculation, never the model's guess.
Web chatLive
A branded, embeddable widget and a hosted chat page. Replies stream in; results render inline.
VoiceBuilt, in rollout
Voice is built but not switched on yet; in-browser and phone voice open together after a live validation pass.
One behaviour, both channels
Chat and voice run the same shared code path, so answers match whichever way a client engages.
Inbound only
The Concierge is inbound only. It responds to interactions your clients start on web chat today (and by voice once that channel opens), and captures the enquiry and consent so your follow-up can continue. It does not initiate outbound calls, emails, or SMS itself; that is the roadmap outbound BDR agent.
The problem
You cannot be on every channel at 2am. The lead can.
Prospects keep shopping between the first click and the first call, and plenty of them land after hours, when nobody is there to answer. A contact form and a callback promise engage nobody while the client fills in three other forms and moves on.
The Concierge is the after-hours cover and the research-window answer: a competent worker on web chat that turns late-night visitors into qualified, captured enquiries, without a broker on the line.
What it does
One worker across the whole top of the funnel.
The Concierge composes what would otherwise be several separate tools, from a curious visitor to a qualified, booked lead. It is thin by design: it decides which feature to call, never re-implementing a calculation or a guardrail in a prompt.
Calculate
It runs the real borrowing-capacity and repayment engines and shows the result as a card. The model chooses to show a calculator; it never authors the figures.
Look up indicative loan products
It runs the real matcher built for your industry and shows an indicative, ranked view of loan products from real lender data.
Qualify and guide
It collects the facts the matcher needs conversationally, one question at a time, rather than forcing a form.
Capture the enquiry
When the client is ready, it writes a real enquiry and its qualification snapshot through the same path the Scout funnel uses, then routes it to a broker through your existing allocation rules.
Take a booking request
It offers the allocated broker's real availability and takes a booking request through a governed show-slots-then-book flow.
Depends on a connected calendar; live connect-and-book is awaiting validation.
Hand off to a broker
When a client is serious or asks for something it should not handle alone, it packages the conversation context and hands the enquiry to a broker.
Captures and routes context; connected live-call transfer is a rollout step.
Feed follow-up
Capturing the enquiry and consent feeds your follow-up automation so an interested client keeps moving. It hands off to that automation; it does not run outbound campaigns itself.
Channels
Chat and voice, one behaviour definition.
A voice call and a chat session resolve the same tools, the same compliance rules, the same persona, and the same knowledge. Voice is not a forked implementation that can drift; it is the same worker with a different channel.
Live
Web chat, fully live.
An embeddable widget for your own site plus a hosted chat page on your subdomain. Replies stream in as the agent thinks, and rich results, a calculator card or a ranked loan-product leaderboard, render inline. A refreshed tab resumes the same conversation.
Built, in rollout
Voice, at capability parity.
Voice is built at capability parity on LiveKit, calling the exact same features as chat and speaking results back as concise summaries. When enabled it opens every call with a spoken AI-processing disclosure, then converses, calculates, qualifies, and takes a booking request by voice.
It is off at launch: both in-browser and live phone voice are gated off until a live validation pass, and the server does not accept a voice session yet.
It cannot make up numbers
Every figure comes from a feature, not the model.
It cannot invent a repayment, quote a made-up rate, or give regulated advice, because those are not things it is allowed to say. They are governed features it must call, and the numbers a client sees always come from the real calculation.
Every AI session carries a model-selection stamp, and an append-only ledger records every model call, so a run can be audited and replayed.
The compliance boundary
It stays in a pre-advice lane, and you stay accountable.
- There is no "advise" feature to call, so the Concierge cannot take a regulated action.
- Deterministic guards force your industry's disclaimer and a handoff offer when a client seeks advice, and an output filter redacts regulated-advice phrasing.
- When voice is enabled, the AI-processing disclosure is spoken verbatim and is guaranteed to finish.
- The broker stays the accountable decision-maker; the AI hands off rather than ever advising.
Live now, and in rollout
What is switched on for you now, and what is coming.
Your web chat is fully live. Voice, booking, and live handoff are built and moving through rollout, not yet switched on for every brokerage.
Voice: in rollout
Voice is built at capability parity with web chat but is off at launch. Both in-browser and live phone voice are gated off until a live validation pass, and the server does not accept a voice session yet.
Booking: ready to validate
Booking is a governed show-slots-then-book flow that depends on a broker having connected a calendar. Calendar scheduling is flag-off by default and awaiting a live validation pass, so it is ready to validate rather than switched on for every brokerage.
Handoff: context, not a warm transfer
Handoff to a broker packages the full conversation context and routes it. It is not a live warm-transfer guarantee: connected live-call transfer, patching a caller straight through to a person, is a rollout step, not a promise yet.
A separately subscribed AI agent
The Client Concierge is a separately subscribed AI agent on its own Starter, Growth, or Pro tier, independent of any tool subscription. Its capabilities are tier- and flag-gated, so what is switched on depends on the plan.
Pricing
A separate subscription, at its own tier.
The Client Concierge is a standalone AI agent, sold on its own Starter, Growth, or Pro tier, never requiring a Spark or Scout subscription. Its capabilities are tier- and flag-gated. See how the tiers compare.
StarterGrowthPro
Client Concierge - the inbound AI agent
Put a 24/7 worker on the edge of your funnel.
The Concierge answers, calculates, qualifies, captures, and takes a booking request, under your brand, on web chat today, with voice built at capability parity for a later rollout. Start free, or book a demo to see it run.