AI
NYE Ride Rush: AI Booking Agents for Burnaby Limo Firms
The 10 p.m. problem
If you run a limo or shuttle company in Burnaby, you know how New Year's Eve goes. The phone starts ringing in early December and picks up through the last week. Then between 10 p.m. and 2 a.m. it gets chaotic: groups change plans, drivers run late and everyone wants a ride home at the same minute.
Most small operators get through it with one dispatcher, a cell phone and a lot of voicemail. Voicemail is where bookings go to die. A caller who can't reach you just tries the next company on the list.
This post covers four things. What an AI booking agent can actually do on that night, what today's tools support, which BC rules you can't hand off to software, and when it isn't worth the trouble.
What an AI booking agent actually is
Without the marketing, it's a voice or chat program connected to your phone line and your booking system. It answers the call and asks the questions a dispatcher would. Then it checks availability, quotes a price, and either books the trip or passes the caller to a person.
The technology exists today. OpenAI's Realtime API guide describes speech-to-speech voice agents where "the model works directly with audio, maintains conversation state, and can call tools." That last part matters. Tool calls are how the agent checks your calendar or writes a booking instead of just chatting.
For phone lines, OpenAI's SIP documentation lays out an inbound flow. Your phone carrier sends the call to OpenAI, your server gets a webhook, and you accept or reject the call. The docs also cover a refer endpoint for transferring a live call to another number.
Two caveats. First, the docs don't describe a built-in human handoff, so your own backend has to decide when to transfer. Second, as of November 2026 the docs are partway through a change from "Realtime" to newer "GPT-Live" names and endpoints. Anyone building on it should expect some changes along the way.
OpenAI isn't the only option. Several vendors offer similar voice tools. We cite OpenAI here because its documentation is public and specific.
What it costs to run (and what we can't tell you)
As of November 2026, OpenAI's API pricing page lists gpt-realtime-2.1 audio at $32 per million input tokens and $64 per million output tokens. The smaller gpt-realtime-2.1-mini is listed at $10 input and $20 output per million audio tokens. Those are model usage prices only. They don't cover your phone carrier, your booking software, or the hours it takes to set up and test.
We won't turn that into a per-call number. It depends on how long calls run, how much the agent talks and how much of each conversation can be cached. Measure it on your own calls in a pilot before you trust any estimate, ours included.
The rules you can't automate away
In BC, limo and shuttle operators can't set whatever prices they like. The Passenger Transportation Board says "The Board sets the minimum and maximum rates operators may charge," and that operators "must tell customers what the rates and charges will be before the trip starts." The Board's limo and shuttle licence page adds that operators "may only charge rates approved in writing by the Board."
That shapes how the agent has to work. It should never make up a price. It should read from a rate table that matches what the Board approved for your licence, and it should state the rate before it confirms a booking. If someone asks for a New Year's premium that isn't in your approved rates, the answer is no, whether a person or software gives it.
Fleet size is also a licence term. The same Board page says operators can apply for a Temporary Operating Permit to add vehicles for up to 92 days, including a Peak Season permit. An agent that books 40 trips when you have 12 licensed vehicles isn't helping you. The booking system has to enforce capacity as a hard limit. A line in the AI's instructions isn't enough.
Privacy: tell people it's AI
In December 2023, Canada's federal, provincial and territorial privacy regulators, including BC's, jointly published Principles for responsible, trustworthy and privacy-protective generative AI technologies. A few of them apply directly to a booking line:
- Openness. Callers should know they're talking to a generative AI tool. One line in the greeting covers it: "You've reached our automated booking assistant."
- Accountability. The principles say accountability "rests with the organization, and not with any kind of automated system." If the agent gets a pickup address wrong, that's your mistake, not your vendor's.
- Limiting collection. Collect only what the trip needs: name, phone number, pickup, drop-off, time and passenger count. You don't need a birthdate to book a ride home from a party.
- Safeguards. The document names prompt injection attacks as a threat. In practice, that's a caller trying to talk the agent into a discount, or into reading out someone else's booking. Keep customer records behind strict tool checks so the agent can't freely browse them.
A sensible New Year's Eve setup
Start with overflow, not replacement. Your dispatcher answers first, and the AI picks up only when the line is busy or after hours. That alone catches many of the calls that used to go to voicemail.
Keep the agent's job narrow: prebooked trips with a fixed pickup time, an approved rate and a confirmation text. Anything messy goes to a person or a callback queue. Say a group of 14 wants a stop in Vancouver, another in New Westminster, then a drop in Burnaby. That's a call for your dispatcher.
Test in early December. Put real calls through it, listen to the recordings, and fix the scripts before the 31st. NIST's Generative AI Profile (NIST AI 600-1) was published July 26, 2024 as a companion to its AI Risk Management Framework. It's voluntary, but it gives you a good way to think about this: identify the risks, measure them, then manage them. For a limo company, that means a short list of ways a booking can go wrong and a test for each one.
Where this doesn't apply
AI booking agents aren't right for every operator, and it's better to say so up front.
If you get a handful of New Year's calls and they're all regulars, a clear voicemail greeting and a callback by 10 a.m. is cheaper and friendlier.
If your booking system has no API or calendar connection, the agent can't check availability. You'd end up with a polite message taker. That might still be worth having, but it isn't a booking agent.
Live dispatch is a different job. Rerouting drivers at 1 a.m. because a client's party moved across town takes judgment, attention to safety and good relationships with your drivers. Keep a person on that.
The tools are still changing. The OpenAI docs we read show endpoints and event names being renamed, with older events still sent to existing setups. That's normal for a young platform, but it means ongoing upkeep. You can't set it up once and forget it.
Voice models also still mishear. Street names, unit numbers and accents on a noisy party line are hard. Have the agent read the address back, and send a text confirmation every time.
The bottom line
On the busiest night of the year, the main win is simple: fewer missed calls, faster confirmations, and a dispatcher who's free for the problems only a person can solve. It works when the agent sticks to your approved rates and real capacity, tells callers it's AI, and hands off cleanly when it should.
Sources
- Office of the Privacy Commissioner of Canada and provincial and territorial privacy regulators, Principles for responsible, trustworthy and privacy-protective generative AI technologies, 2023. priv.gc.ca
- National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1), 2024. nist.gov
- OpenAI, Realtime API developer guide, accessed November 2026. developers.openai.com
- OpenAI, Voice SIP developer guide, accessed November 2026. developers.openai.com
- OpenAI, API Pricing, accessed November 2026. developers.openai.com
- BC Passenger Transportation Board, Limo and other PDV rates, accessed November 2026. ptboard.bc.ca
- BC Passenger Transportation Board, Limo and shuttle licence, accessed November 2026. ptboard.bc.ca
Thinking about an AI booking agent before New Year's Eve? Autana Solutions is based in Burnaby and New Westminster, and we'll give you an honest answer, including if your business doesn't need one yet. Book a free call and we'll look at your call volume, your booking system and your licence terms together.
Want an AI employee for your business?
We install a 24/7 AI worker for businesses in Vancouver, Burnaby, and beyond. Book a free Discovery Call.
Book a call →

