The fastest way to catch a misconfigured greeting, missing service, or broken transfer is to talk to your AI before patients do. A test call lets you do exactly that — your AI calls you, using whichever script you pick, so you hear the real thing.
Start a test call
- Open Settings → AI Voice → Call scripts.
- Find the script you want to try. It must be enabled first — the Test button is greyed out for disabled scripts.
- Click Test on that script's row.
- Enter the phone number you want the AI to call. Click Use my phone to prefill your practice number.
- Click the green call button. Your phone rings within a few seconds — answer it and talk to your AI.
The call uses that script's exact prompt plus your current voice from the Voice tab, so it reflects your live setup. If you change the voice on another tab and re-test, the new voice is picked up automatically.
What happens during the call
The modal walks through each stage:
| Stage | What you'll see |
|---|---|
| Connecting | The AI is setting up the call. |
| Ringing your phone | Your phone is ringing — answer it. |
| Call in progress | A live timer and a transcript of the conversation as it happens. |
| Complete | A success screen with the full transcript. Click Test Again to rerun. |
| Failed | A troubleshooting screen (see below) with retry options. |
Because test calls run in interactive mode, the AI has access to your real tools — booking, knowledge base, transfers — with simulated results. That means you're validating genuine behavior ("did it find an open slot and confirm?"), not just hearing it read a script.
What to listen for
While you're on the call and when you review the transcript, check:
- Greeting accuracy. Practice name, and anything time-sensitive (closed for a holiday?).
- Service recognition. The AI matched your wording to the right service ("cleaning" → "Adult Cleaning", "checkup" → "New Patient Exam").
- Provider names. Pronounced acceptably.
- Transfer behavior. Right destination and the right wording before the hand-off.
- Clean sign-off. Confirmation read back, polite goodbye, no awkward silence.
- Refusals. Refill, medical question, or reading a card number — the AI declines correctly and offers an alternative.
Test-call allowance
Test calls draw from a per-plan allowance, shown in the modal. Lower plans include a set number of test calls; higher plans include unlimited. If you reach the limit, you'll see an upgrade prompt instead of a new call.
If a call fails
The failure screen names the likely cause (busy, no answer, declined, invalid number) and suggests fixes:
- Make sure your phone is on and has signal.
- Check the number for call blocking or spam filtering.
- Try a different number — mobile numbers connect most reliably.
Use Try Again to redial the same number or Use Different Number to change it.
Fix what you find, then edit the script
When a test surfaces the wrong behavior, adjust the underlying instructions on the Call scripts tab — or capture the correction as a permanent rule in the Review queue and clinical rules. Then test again.
Saved scripts run from the same screen with one click.
Recommended cadence
| When | What to run |
|---|---|
| Before going live | All 6 built-ins + 2 of your own real-world scenarios |
| After adding a service | New appointment + your custom script for that service |
| After changing greeting | New appointment + General inquiry |
| After changing FAQ knowledge | General inquiry covering the new topic |
| After changing transfer rules | Prescription refill (or whichever trigger maps to the new rule) |
| Weekly health check | All 6 built-ins (takes ~3 minutes) |
Replay a real call
Open any call from Calls → Logs, click ⋯ → Replay as test, and the transcript becomes an editable script. Tweak settings to fix the problem, re-run, and confirm the behavior is right before the next real caller sees it.