Zenith agents
One front desk, and the right agent for each conversation.
A front desk that knows everything about everyone is hard to build and harder to trust. Zenith splits the work instead: one inbound agent, Front desk, sits on every call and thread, and brings on a narrow agent for the job in hand, whether that is a reservation, a catering inquiry, an order question or a recall. Each sees only what its job needs, and your rules decide which one comes on.
Pre-built
Six agents every company starts with.
Each is configured to your business when your instance is set up, and each has a plain answer for what it does on its own.
- Front desk
- Your inbound agent. On every call and thread: identifies the caller, loads the profile, brings on the right agent, writes the record.
- Lead responder
- New people on any channel, qualified with your questions and booked into the next step.
- Scheduler
- Appointments and reservations booked, moved and reminded in your own calendar or system.
- Customer care
- Hours, order status, policies and menu questions answered from your records; everything else taken down for a person.
- Reviews
- Requests after a visit, a drafted reply to every review, low ratings held for a person.
- Win-back
- Lapsed customers found in your records each night, with drafts for a person to approve.
On a call
How Front desk picks the agent, one at a time.
Selection is your configuration, evaluated in code, first match wins. Front desk's own judgment is used only to recognize what the caller is asking for.
- 01
Who is calling
Front desk matches the caller: a known customer, a known lead or a new person, and loads the profile.
- 02
What is happening
Triggers from the profile and the line: a reservation tonight, an open order, a recall due, the catering number, the word manager.
- 03
Who comes on
Your selection rules name one agent. Its narrow context loads, and it guides the conversation through Front desk.
- 04
A new topic
If the caller asks for something else (can I also book a table?), Front desk swaps agents and hands over a summary.
- 05
The record
One record per call: which agents joined, why, and what each one did.
Custom agents
As many agents as your business needs.
The six are the starting set. Companies add their own contextual agents for the jobs particular to them, each built by picking options, never by chat.
A large-party desk
For restaurants: date, size and occasion, capacity by night, and a tasting booked with the events lead.
A recall desk
For dental: who is due, which provider, and times that suit the patient's past visits.
A warranty desk
For home services: the equipment on file, its warranty status and the right technician.
An order desk
For distributors: open orders, ship dates and the account rep, read from the ERP.
Takeover
A person can take over any conversation.
| While an agent works | When a person takes over | |
|---|---|---|
| A live call | You can listen in from the call screen | Transfer to yourself or a colleague; the caller's context goes with it |
| A text or chat | Replies are labeled with the agent's name | The agent stops at once and leaves a summary |
| A handoff | The agent decides a person is needed and says why | The thread moves to Needs a person with any draft prepared |
| Afterwards | The agent follows up on its own | You close the thread or hand it back |
On its own, and where it stops
Zenith handles the routine. Your people handle the rest.
Agents answer, route, qualify, book and handle routine questions on their own. Refunds, discounts, complaints, low reviews and campaigns are drafted or taken down for a person, and nobody is ever more than one click from taking over.
Zenith answers, routes, books and handles routine questions on its own. Refunds, discounts, complaints, low reviews and campaigns wait for a person, and anyone on your team can take over a conversation at any time.
Your dataThe voice model carries the spoken conversation and nothing else. Your customer records stay in your own database, read only by your own agents.
Start here, keep the foundation
Start with one function. Add the rest without starting over.
Most companies start with Front desk and one contextual agent for the call they get most, often scheduling. Every agent added after that reads the same caller profile and the same records, so each new one is a set of rules and context, not a new integration.
FAQ
Questions, answered plainly.
What are AI agents for customer service?
They are software agents that hold conversations with customers and do the work those conversations call for: answering questions, booking, changing orders or routing to a person. In Zenith, each agent has one job, its own narrow context and only the tools that job needs.
Why one agent at a time instead of one agent for everything?
A narrow agent is easier to configure, test and trust. Front desk guides the voice assistant through exactly one agent at a time, so the caller never gets two sets of instructions, and every swap is logged with its reason.
Who decides which agent comes on?
Your selection rules, evaluated in code, first match wins. Front desk names the caller's reason when they state it, and if two rules both allow an agent, picks between them. It can never bring on an agent your rules do not allow.
How do we build a custom agent?
By picking options: what it handles, which records it may see, which tools it may use, which rule brings it on and when it hands to a person. There is no chat-built agent. Ferrata builds the first ones with you and your team builds the rest.
What does an agent never do on its own?
Issue a refund, give a discount beyond your limit, settle a complaint, post a reply to a low review or send a campaign. Those are drafted or taken down and wait for a person.
Can our team see what an agent did and why?
Yes. Every conversation records which agent worked it, which rule brought it on, what it said and what it changed in your systems. Swaps and handoffs carry their reasons.
Do agents work on texts and chat as well as calls?
Yes. The same Front desk and the same agents work every channel. On a call they guide the voice assistant; on a text, chat or email they write the reply.
Related in Zenith
- AI receptionistEvery call answered by a voice that knows the caller, books the visit and hands off with context.View
- Call routingEvery call sent by the line dialed, the time, who is calling and why.View
- Lead responseEvery new lead answered on the channel they used, qualified and booked, day or night.View
- Appointment schedulingAppointments and visits booked into your real calendar or system, confirmed and reminded.View
- Customer reactivationLapsed customers found in your POS or ERP, with offers drafted for a person to approve.View
From Ferrata Labs
- Ferrata's agentsSix agents, each named for the operational work it does.
- The agent lifecycleScope, build, evaluate, gate, deploy, observe, improve.
- The platformThe Brain: retrieval, models and audit inside your own environment.
- The embedded teamTwo forward-deployed engineers, working with your IT team.
- Book a discovery callThirty minutes on one workflow, and an honest answer on where to start.
Start with Zenith, or talk it through first.
Thirty minutes with Ferrata Labs on how calls and messages reach you today. You leave with a view of where Zenith fits, and an honest answer if something else should come first.
Your menu, your policies, your routing and the way you talk to customers are yours. Zenith is configured to them. You are not fitting your front desk into a template.