Voice Agent for Big Dealership with 2000+ calls per day, with no API to their system

I am currently working on a project involving an AI voice receptionist for a network of car dealerships operating across the country.
There is one central phone number, and all incoming calls are handled through this main line. The company has multiple dealership locations nationwide, and currently a human customer service team (BOK) answers the calls, identifies the correct location, checks availability in the service calendar, and manually enters appointments into their internal system.
The goal is to automate this process as much as possible.
The AI receptionist would:
answer calls,
identify which location the customer is referring to,
collect necessary information (name, phone number, VIN or registration number, type of service),
and most importantly, book the appointment directly into the service calendar.
The volume is significant — approximately 2,000 calls per day.
The main priority for my client is very clear:
they want the appointment to be entered directly into their existing service calendar system, completely bypassing the human customer service team.
However, the challenge is that their internal service system does not provide an API, and there are no plans to introduce one in the near future. There is no official integration mechanism available.
We discussed a few possible approaches.
Initially, I considered creating an intermediary database (for example, a central scheduling database) where the AI would store bookings and then notify the appropriate location via SMS or email. However, this does not fully eliminate the need for human involvement, and the client strongly prefers a solution where the appointment is written directly into the service calendar.
Another idea discussed was using login credentials to access their system — essentially automating the process by logging into the service platform the same way a human would and entering the appointment automatically.
So the key question becomes:
Is using system logins (UI automation / browser automation) a realistic and stable solution at a scale of 2,000 calls per day?
Would such an approach make sense in a high-volume environment?
Would session limits, timeouts, security policies, or concurrent operations become a major issue?
Is there a more stable way to achieve full calendar insertion without an official API?
The client is primarily focused on completely removing the BOK team from the booking process, and direct calendar entry is their main requirement.
Given the constraints — no API, high call volume, and a strong need for full automation — I would really appreciate insight into what would be the most robust and scalable way to approach this.

Information on your n8n setup

  • n8n version: 2.7.4
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app): Railway
  • Operating system: Windows 10

The Hard Truth About UI Automation at Scale UI automation (browser/RPA) at 2,000 calls/day is risky and generally not recommended as a primary solution for several reasons:

  • Fragility: Any UI change breaks your automation

  • Speed: Browser automation is slow (10-30+ seconds per booking)

  • Concurrency limits: Most systems have session limits (5-10 concurrent logins typical)

  • Detection/blocking: Many systems detect and block automated logins

  • Error handling: Failed bookings are hard to recover from

  • Maintenance burden: Constant monitoring and fixes required

At 2,000 calls/day (~83/hour during business hours), you’d need multiple concurrent sessions running continuously. This is where UI automation typically falls apart.

Best option : Hybrid Approach (Most Practical)

AI handles call → Structured data to queue → Reduced human team processes queue

  • AI collects all information perfectly structured

  • Writes to a queue/database with all details pre-validated

  • Small BOK team (much smaller than current) processes queue rapidly

  • They only do data entry, no customer interaction

Why this works:

  • Reduces BOK team by 70-80% (no phone time, just fast data entry)

  • Reliable and maintainable

  • Client may resist initially, but it’s the most stable solution

  • Can add UI automation later as a “helper” for the reduced team

At 2,000 calls/day, you need enterprise-grade reliability. UI automation alone won’t give you that. The client’s desire to “completely bypass BOK” is understandable, but without proper integration points, you’re building on sand.

The best path forward: Sell them on the hybrid approach as Phase 1, then work toward proper integration. Show them that 80% automation with 100% reliability beats 100% automation with 60% reliability.

Does this help!

1 Like