TryGTM is live.See it

Plan for the day the booking system stops answering

7 min read

By Lumina Software

  • Voice reception
  • Service business
  • Reliability
  • Operations
  • Scheduling

A phone agent that books jobs is only as good as the systems behind it. When it answers a call, it usually reaches into a calendar to find open times, a customer record to look up the caller, and a job system to create the visit. Each of those is a separate piece of software, often from a separate vendor, connected over the internet.

Most days, all of it works. Then one morning the scheduling software has an outage, or a password expires, or the vendor changes something on their end. The phones keep ringing. The agent keeps answering. And the question nobody planned for shows up on a live call: what does the agent do when it cannot see the calendar?

If nobody wrote that down, the agent will do something anyway. It might tell the caller it booked them when it did not. It might apologize in a loop. It might hang up. None of those are good, and the caller has no way of knowing that the problem is on the back end.

This post covers how to give the phone agent a written plan for the day its tools stop answering.

Know which tools the agent depends on

Start with a short list. For each step the agent takes on a typical call, write down what system it talks to:

  • Looking up the caller: the customer database or CRM
  • Finding open times: the scheduling or dispatch calendar
  • Creating the job: the field service or job management system
  • Sending the confirmation text: the messaging provider
  • Handing off to a person: the phone system and the on-call schedule

Next to each one, write what the agent should do if that one step fails. This sounds like extra work, but it is usually a half page. Most businesses only have three to five systems in the call path, and the failure plan for each is a sentence or two.

The important part is that the plan is per step. A failed customer lookup is a small problem. The agent can ask for the caller's name and address and carry on. A failed calendar is a bigger one, because the agent can no longer promise a time. Treating every failure the same way, usually by ending the call, throws away calls that could have been handled fine.

Never confirm what you could not save

The single most damaging failure is a booking the caller thinks is real and the business has no record of. The caller takes the morning off. Nobody shows up. The business finds out when the customer calls back angry, or does not call back at all.

The rule that prevents this is simple: the agent only says "you are booked" after the job system has confirmed that the visit was saved. Not after it sent the request. After it got an answer back saying the record exists.

If the save fails or times out, the agent should say something like "I have all your details, but I was not able to lock in the time on our schedule. Someone from our office will call you to confirm." Then it has to make sure that is true, which is the next section.

This rule also matters when the system is slow instead of down. A save that takes thirty seconds may still go through. If the agent retries, it should check whether the first attempt landed before it creates a second one, or the business ends up with two visits for the same caller.

Capture the call even when you cannot book it

When a tool is down, the goal shifts from booking the job to not losing it. Every call should leave behind a complete record somewhere the team will see it, even if the main job system is unreachable.

That means the agent needs a fallback place to write. It should be something simple and separate from the systems that just failed. A shared inbox, a text to the office manager, or a plain spreadsheet all work. What matters is that it does not depend on the same vendor as the calendar.

Each fallback record should have:

  • The caller's name and callback number
  • The address and what the job is about
  • The time window the caller asked for
  • Which step failed, so the office knows why it landed here
  • A clear flag that this caller was told someone would call back

The last item is what turns a pile of notes into a to-do list. When the office opens the fallback inbox, every entry should read as a promise the business made, with a name and number attached.

Tell the caller something true

Callers are forgiving about problems. They are much less forgiving about being misled. The script for an outage should be honest, short, and specific about what happens next.

Write the exact wording for the two or three situations that matter most, and have the owner approve it:

  • Calendar is down. "Our scheduling system is having trouble right now, so I cannot pick a time with you on this call. I have your details, and someone from our office will call you by [time] to set it up."
  • Job system will not save. "I was not able to finish the booking on our end. I have everything you told me, and our office will call you to confirm the time."
  • Everything is down. "I am having trouble reaching our systems. Let me take your name, number, and what you need, and I will make sure a person calls you back."

Notice that none of these blame the caller, guess at the cause, or promise a fix time the business cannot control. They tell the caller what the agent has and who will follow up.

If the business has an urgent or emergency path, the outage plan should not touch it. A caller with a real emergency still gets routed to the on-call person by phone, whether or not the calendar is working. Keep that path as independent of the booking tools as you can.

Find out before the callers do

An outage plan works best when someone at the business knows it is running. If the agent falls back on three calls in a row, that is not three bad calls. It is one broken system, and the office should hear about it right away.

Set up an alert that fires when fallback records start piling up. A simple threshold is enough: two or three fallbacks within an hour sends a text to the owner or office manager. The message should say which step is failing, so the person reading it knows whether to call the scheduling vendor, check a password, or just start working the callback list.

It also helps to test the plan on purpose. Once a quarter, during a slow hour, turn off the agent's access to the calendar in a test setup and place a call. Listen to what it says. Check that the fallback record arrived with all its fields. Then turn access back on and confirm the agent recovers without anyone touching it. A plan that has never been run is a guess.

Work the callback list the same day

The fallback only helps if someone closes the loop. Every caller who was told "someone will call you" needs that call, ideally within the window the agent named.

Give the callback list one owner each day. When the outage ends, that person books each fallback job in the real system, calls the customer to confirm the time, and marks the entry done. If a caller cannot be reached, leave a voicemail and send a text with a number to call back.

At the end of the week, look at how many calls went to the fallback and why. If one system shows up again and again, that is worth a conversation with the vendor or a change in how the agent connects to it.

The short version

A phone agent depends on several outside systems, and sooner or later one of them will stop answering in the middle of a call. List each system in the call path and write what the agent does when that step fails. Never tell a caller they are booked until the booking is saved. Send every failed call to a simple fallback that does not share a vendor with the broken tool. Give callers honest, approved wording about what happens next. Alert someone when fallbacks start stacking up, and make sure every promised callback actually happens.

Written by Lumina Software. Have a question? Book a call

Keep reading