ai-receptionist,buyer-guides
Call routing rules for multi-department businesses
A single-location business with one phone line has it easy. The AI answers, handles the call, done. But the moment you have departments, a service team and a sales team, a front office and a back office, the AI needs routing rules: decisions about which calls go where. Get routing wrong and callers bounce between departments explaining themselves twice. Get it right and the AI becomes the switchboard operator you never had to hire.
Map departments to caller intents, not to your org chart
The routing map should be built from the caller’s perspective, not yours. Your org chart might have “operations,” “client success,” and “dispatch.” Your callers think in “I need to book something,” “something’s wrong with what you did,” and “I have a question about my bill.” Build the routing around those intents.
For each department, write down in plain language what it owns:
- Sales / new business: first-time inquiries, quotes, estimates.
- Scheduling / dispatch: bookings, reschedules, “when are you coming?”
- Service / support: problems with completed work, warranty questions.
- Billing: invoices, payments, disputes.
- General / front office: everything else, plus the fallback when the AI can’t classify the call.
The AI’s job is to listen to the caller’s first description and match it to one of these. It doesn’t need your internal team names; it needs the intent buckets. Train the routing on examples of how callers actually describe things, because nobody calls and says “please connect me to client success.” They say “the thing you installed last month stopped working.”
The routing questions that disambiguate
Most calls route cleanly from the caller’s opening sentence. The tricky ones need one clarifying question. Build these into the routing rules:
New vs. existing: “Are you a current customer, or is this your first time calling us?” This single question splits the map in half. New callers go toward sales and booking; existing callers go toward service, scheduling, or billing.
Urgency: “Is this something that needs attention today?” An urgent service issue routes to dispatch or the on-call person. A non-urgent one can be a message or a next-available booking. This question belongs in your escalation rules too; routing and escalation overlap here.
Department confirmation: When the AI isn’t sure, it should say where it’s sending the call and why: “Sounds like a billing question, let me get you to our billing team.” If the caller corrects it, the AI re-routes. Stating the destination out loud catches misroutes before the transfer instead of after.
Direct, queue, or message: the three routing outcomes
Every routing decision ends in one of three places. Configure all three per department.
Direct transfer is for the calls that need a person now: urgent issues, hot sales inquiries during business hours, the caller who asked for someone by name. The AI transfers with a summary so the receiving person is briefed.
Queue or hold for the next available person works when the department is staffed but busy. The AI should tell the caller what’s happening: “Let me get someone from scheduling. It’ll be just a moment.” If the wait stretches past a couple of minutes, the AI should offer the alternative: “They’re tied up right now. I can take your details and have them call you back within the hour, or I can keep holding. What works better?”
Message taking is for after hours, unstaffed departments, or callers who prefer it. The message needs the same five essentials every time: name, number, what it’s about, how urgent it is, and callback preference. A message missing the urgency flag is a message that sits until someone gets around to it.
The fallback department
Some calls won’t fit any bucket. The caller rambles, the request spans two departments, the AI genuinely can’t tell. Every routing map needs a fallback: a person or team that receives the unclassifiable calls. Usually this is the front office or the owner. The AI’s line for this should be honest: “I want to make sure you get to the right person. Let me connect you with our front office and they’ll sort it out.”
Track how often the fallback gets used. If it’s more than a small fraction of calls, your intent buckets are wrong and need reworking. The fallback is a safety net, not a department.
Routing across locations vs. routing across departments
If you have multiple locations as well as departments, keep the two routing decisions separate. First the AI determines which location (usually by the caller’s address or the number they dialed), then which department within that location. Mixing them into one decision produces a routing table nobody can debug. If multi-location routing is your bigger problem, sort that first; the multi-location call routing setup covers it, and department routing layers on top cleanly.
Write the routing map on the same page as your call flow map. Departments, intents, the disambiguating questions, and the three outcomes per department. One page, reviewed monthly against the transcripts. Routing is the part of the setup that touches every call, so it’s worth the hour it takes to get right.
I write the agents that run on clinic phone lines and inboxes: the conversation engine and the booking logic behind them, plus the integrations with Pabau, Fresha and Phorest. Everything here comes out of systems we have actually shipped, not a content plan.
