AI Agent for Multi-Branch Networks: Route Leads Right

Thursday, 17:40. A dental network with four branches runs one campaign for the whole chain. A woman in Petah Tikva writes to the Instagram page: "how much are clear aligners and when do you have an opening?". The message lands in the head office inbox, the marketing coordinator screenshots it into the branch managers' group asking "whose is this?", and the Petah Tikva manager sees it Sunday morning. By then she has booked elsewhere. The network lost that lead not for lack of an answer, but because the inquiry arrived at the network while the service happens at a branch.

Why a network loses more leads than a single location

A single-location business has one inbox, one calendar and one person responsible. In a network all three split: the inquiry arrives at one digital asset (the chain website, a Facebook page, a WhatsApp number, a campaign), but the service happens at a specific branch with its own calendar and staff. In that gap sits a role nobody defined: the dispatcher, in practice the marketing coordinator or the owner, working office hours. Three costs follow:

• Latency. Hours pass between the customer writing and a branch replying, sometimes a weekend. Someone looking for treatment messages three places and picks whoever answered first.
• Leakage. A screenshot in a WhatsApp group is not a system. Inquiries get swallowed by the scroll, and nobody knows how many, because what is not recorded appears in no report.
• Blindness. With no branch-level record you cannot tell which branch closes well and which one leaks, so the network is run on gut feel.

Three ways to handle inquiries in a network

A separate number per branch. No dispatcher, but the brand splits into four numbers, every campaign needs its own destination, and nobody at head office sees the full picture.

One number with a human dispatcher. One brand and an internal queue somebody forwards by hand. Works until that person is on holiday, in a meeting, or it is 21:00.

One channel with an AI agent that routes. A single entry point and automatic routing: identifying the branch, answering from a shared knowledge base, booking that branch's calendar and opening a card in its name. Anything needing a person goes to the branch team with the conversation history. Hybrid by design: the agent takes the volume and triage, the branch takes the judgement.

How branch routing actually works

Routing is not magic but a rules table written during setup. Four signals, in priority order:

1. What the customer wrote. "I am in Ramat Gan", "the Netanya branch". The strongest signal, usually present in the first or second message.
2. One clarifying question. With no information, the agent asks a short question listing the branches ("which is convenient: Petah Tikva, Rishon LeZion, Netanya or Haifa?"). One question, not a form.
3. The source of the inquiry. A city-targeted campaign, a branch page, a dedicated WhatsApp button or a QR code inside the branch. Those parameters sit on the card and set a default before anything is asked.
4. Calendar availability. When the requested branch is booked two weeks out and its neighbour is free tomorrow, the agent offers both by a rule the network defines. A business decision written in advance, not a technical default.

Two supporting rules prevent most failures: when information is missing, ask rather than guess, because a wrong assignment costs more than one extra question; and a handoff goes to the branch team rather than head office, which is the path we wanted to delete. How consolidating channels into one inbox changes response times is covered in our unified inbox guide.

One knowledge base with local exceptions

The classic mistake is one knowledge base that fits every branch "roughly". The result shows up fast: the agent promises a treatment the Haifa branch does not perform, or quotes another branch's hours.

The structure that works separates two layers. The shared one holds what is identical across the chain: service catalogue, price list, cancellation policy and the brand's tone of voice. The local one holds, per branch, what genuinely differs: address and parking, opening and holiday hours, which services that branch performs, and who receives an urgent handoff. Holiday hours cause real pain in Israel: holiday eves, Chol HaMoed and fast days differ at every branch, and without a local layer the agent books a slot when the branch is closed.

The calendar is configured per branch, the customer gets a confirmation email with the correct address, and a reminder the day before cuts no-shows. Follow-ups and workflows are unlimited on every plan, with no trigger meter - which matters where each branch runs its own reminder sequence. For clinic networks the full picture is on our AI agents for clinics page.

Who owns the lead, and what the network manager sees

In any network with more than two branches, and all the more so in a franchise, the lead is political: "that customer is mine", "the inquiry came from our campaign", "the branch never got back to her and I closed it". Without a record, the argument is settled by whoever remembers better.

The fix is operational and simple: every conversation opens as a card in WhaleBiz CRM with a branch field, an owner, a source and a timestamp, and reassignment is logged on the card with its reason. A branch manager sees their own inquiries, the network manager sees everything, and a response-time rule escalates anything nobody touched. It works because conversation and card sit in one system rather than two that need syncing - the idea we detailed in our guide to a CRM built around AI agents.

From that comes the report managers feel first: one table with inquiries per branch, how many arrived outside opening hours, how many appointments were booked, and which questions the agent could not answer. The failed questions are exactly what the knowledge base is missing, which makes that the most useful column of the first month.

How to roll this out without breaking the branches

1. Pilot one or two branches. A branch with decent inquiry volume and a manager willing to cooperate. Rolling out to the whole network in one day is the reliable way to create resistance.
2. Freeze the knowledge base. Price list, service catalogue and cancellation policy, and per branch: address, hours, services and the contact for handoffs. The longest stage, and almost all of it internal organising.
3. Write the routing table and the handoff. The four signals, the default rule, and who receives a handoff at each branch and during which hours.
4. Measure for three to four weeks, then expand. Update the knowledge base from the failed questions, and only then roll out further.

Easy to miss: bring the branch managers in at stage two. A branch that sees the agent as a tool sending it ready-made leads will cooperate; a branch that feels monitored will fight it.

What it costs

Solo is 490 ILS per month: one agent on one channel, 2,000 AI messages per month (about 300 conversations), a knowledge base of up to ten documents and WhaleBiz CRM. Pro is 990 ILS per month: up to three channels, 5,000 AI messages (about 750 conversations), an unlimited extended knowledge base, up to a hundred media files and in-chat payments. A small network on one channel starts on Solo, most land on Pro, and a separate number per branch or an integration with an existing system is scoped on Custom. On top of that, a one-time setup fee from 490 ILS covers full setup by our team, quoted by scope; setup takes up to 14 business days from receiving all materials, and the subscription starts counting only when the agent goes live. Prices exclude VAT, and the breakdown is on the pricing page.

The relevant comparison is not against "nothing" but against the dispatcher: half a coordinator's time forwarding inquiries between branches costs far more, and still does not answer at seven in the evening. Routing logic, booking and follow-ups are what the AI sales agent covers.

Frequently asked questions

Does every branch need its own AI agent?

One agent knows all the branches and routes between them: each branch has its own address, hours, calendar, contacts and services inside that agent. That is easier to maintain than an agent per branch, since a price or policy update is made once. A separate WhatsApp number per branch is scoped on Custom.

How does the agent know which branch an inquiry belongs to?

From four signals set in advance: what the customer wrote (a city or branch name), one short clarifying question listing the branches, the source of the inquiry (a city-targeted campaign, a branch page, a specific WhatsApp button) and calendar availability. When information is missing the agent asks instead of guessing.

What stops branches or franchisees arguing over lead ownership?

Every conversation opens as a card in WhaleBiz CRM with a branch field, an owner and a timestamp, and reassignment is logged with its reason. You can see who received the inquiry, when and what was done with it, so there is one record instead of an argument about who remembers what.

What does an AI agent cost for a network with three or four branches?

A small network on one channel can start on Solo at 490 ILS per month (2,000 AI messages, about 300 conversations). Most land on Pro at 990 ILS per month: up to three channels, 5,000 AI messages and an unlimited extended knowledge base. A separate number per branch or an integration with an existing system is scoped on Custom. Every plan adds a one-time setup fee from 490 ILS, and prices exclude VAT.