An AI agent is a worker you hire for a role, not another tool you have to wire up. It answers, qualifies, and captures leads on its own, under your brand. One is live today, the inbound Client Concierge: web chat is live now, and voice is built at capability parity but off at launch. The rest are the direction we are building toward.
The same underlying features can appear inside a tool or inside an agent. The difference is who does the assembling: with a tool, you wire the jobs together; with an agent, the worker composes them into one role and talks to the client in plain language.
A tool such as Spark or Scout does one focused job in the client journey and leaves you to wire the jobs together. Buy the pieces, drop them onto your pages, and connect them yourself. Powerful, but a toolbox you hold.
An AI agent packages every feature its role needs and runs them itself, then talks to the client for you. Subscribe to the Client Concierge and get education, calculations, indicative loan-product lookup, qualification, enquiry capture, and booking in one worker. No stitching Spark, Scout, and Nudge together first.
An AI agent subscription is independent of any tool subscription. Run the Concierge with no Spark or Scout plan; the shared feature code is reused, but the entitlement stands on its own.
Every TomoBroker AI agent runs on one industry-agnostic foundation. Four design choices define it, and each earns a benefit you can point to.
Calculating a repayment, matching a loan product, capturing an enquiry: each lives in a reusable feature built for your industry. The agent owns packaging, persona, and orchestration, and calls the same vetted engine your dashboard uses, so an agent answer and a broker answer agree.
Every run happens inside your own account, under your permissions, usage limits, and audit trail, built for the industry you serve. Nothing the agent does sits outside that model, so the AI behaves like part of your regulated business, not a loose add-on.
The agent plans, converses, and recovers; the deterministic features do the calculations, the matching, the state changes, and the compliance checks. A model cannot invent a rate or a repayment, because those come from a feature, not from its own text.
The same feature path serves web chat at launch, with voice built at capability parity for a later rollout and SMS, email, and API provisioned as future channels on the same runtime. Configure an agent once and its behaviour, knowledge, and guardrails are the same whether the client types or calls.
Being capable is not enough in a regulated market; every answer has to be governed. Per-brokerage isolation, deterministic compliance, a full audit trail, and controlled model selection are built into the foundation, so you can put the agent in front of your own clients.
Every run is scoped to one brokerage. An anonymous public visitor can only create new data; reading or changing an existing client or enquiry needs proven assurance. One brokerage's agent can never reach another's data.
Disclosure, disclaimers, and the advice boundary are enforced by deterministic rules, not left to the model's discretion. The regulated-advice boundary and its disclaimers are built for your industry and country, so they are correct for the market you serve.
Every run records what the agent was asked, which features it offered and called, the facts it relied on, the disclaimers it showed, and the model behind it. Results shown in a card are backed by immutable snapshots, so "what did the client see then?" can always be reconstructed.
The platform picks a vetted model profile per brokerage, plan, and channel and records the choice, rather than letting anyone point the agent at a raw model. A profile can be retired or emergency-disabled fleet-wide and the agent keeps working on a substitute.
The Client Concierge is the one built agent today. Every other role below is roadmap or concept, marked as such, and none is presented as something you can subscribe to yet. When in doubt, lead with the Concierge.
The inbound agent, built and shipped. It answers the conversations your clients start on web chat (live) and voice (built, in rollout), then educates, calculates, qualifies, captures the enquiry, and books. It feeds your follow-up; it does not run outbound outreach itself.
The outbound counterpart to the Concierge. Where the Concierge waits for the client to make contact, this agent starts the conversation: initiating sessions, calls, emails, and SMS to follow up leads and re-engage quiet clients, on the same governed, audited foundation. A distinct role from the Loan Processor, and not available yet.
A back-office agent that prepares files for your review: chasing missing information, checking documents, extracting data, and flagging inconsistencies. It prepares, checks, and flags; it never replaces your accountability for the file. Planned, not yet built.
Not a product you can buy. It is the umbrella term for the whole category of broker-controlled agents that act under your brokerage's permissions and audit model, the inbound Concierge and the roadmap agents alike. Use it for the direction, an AI workforce for your brokerage, not for something available today.
The Client Concierge is industry-agnostic: everything domain-specific is built for the industry you serve. An Australian mortgage Concierge talks about deposit, borrowing capacity, and home loan products; a future life-insurance Concierge would talk about sum insured and policy options, using the same worker.
The same features, permission model, and audit trail carry forward to the outbound counterpart and to industry-specific agents without a rewrite. That is the direction we are building toward, not a spread of live agents today.
The through-line
The inbound Client Concierge answers, calculates, qualifies, captures enquiries, and takes booking requests, under your brand, 24/7. Start free, or book a demo to see it in action.
