How AI Employees Are Trained
How an AI Employee Learns Your Rates, Zones and Policies
It is trained on your own material rather than industry averages. Your rate table, your zones, your vehicle classes, your surcharges and your policies are loaded in, so the price it quotes is the price you would have quoted. Where it is unsure or the question falls outside what it was given, it hands the call to a person instead of guessing.
Key points
- It works from your rate card, your zones, your vehicle classes and your surcharges — not generic figures.
- Your policies go in too: cancellations, waiting time, child seats, what you will and will not do.
- The quote it gives should match the quote you would have given. That is the test.
- Where it is unsure, it hands over to a person rather than inventing an answer.
- Setting it up is a conversation about how your business actually works.
- Review it after the first week of real calls. That is where the gaps show.
The reason most AI quotes are wrong is that they are generic. The system knows what a limousine is; it does not know what your airport transfer costs on a Sunday evening with two suitcases and a child seat.
The fix is unglamorous. Your actual rates, zones, vehicle types, payment methods and procedures are loaded in before it answers a single call.
Why a generic quote is worse than no quote
Consider what happens when the number is wrong.
A caller asks the price for a run to the airport. The system produces a figure that sounds plausible and is $30 under your rate. The caller books at that price.
You now have three bad options. Honour it and lose the margin. Ring the customer back and correct it, which is an awkward call about money before you have driven a metre. Or let the driver sort it out at the door, which is the worst of the three.
Compare that with a system that says "let me have someone confirm that for you and call you back". Less impressive in a demo. Vastly better on a Tuesday.
This is why the loading work matters more than the conversational quality. A system that speaks beautifully and prices wrongly is a liability.
What gets loaded before go-live
Five categories, and each one is drawn from how you already operate rather than from a template.
Your rates. The full card, not a starting price. Covered in detail below.
Your service zones. Where you work, where you charge differently, and where you do not go. Zone boundaries are where generic systems fail most obviously, because they have no idea that the far side of a particular river is a different price.
Your vehicle types. Saloon, executive, MPV, minibus, whatever your classes are called, with the capacity and the price difference between them.
Your payment methods. What you accept, when payment is taken, whether corporate accounts are invoiced, what the deposit policy is.
Your operating procedures. Cancellation terms, waiting time rules, what happens on a no-show, how far in advance you take bookings, meet-and-greet arrangements.
How your fares actually work
Rates are the part operators most expect to be simplified, and it is the part that must not be.
A working rate card usually mixes several methods at once:
- Fixed prices for named journeys you do constantly
- Distance-based rates for everything else, generally with a minimum
- Per passenger for shared or shuttle work
- Zone to zone, a matrix with a price for each pair
- Hourly for as-directed and chauffeur-by-the-hour work
Sitting on top of those are the charges that decide whether the job was profitable: airport pickup and drop-off fees, waiting time beyond the free period, child seats, meet-and-greet, and the windows where peak, holiday or night pricing applies.
All of it gets loaded, because a quote that ignores the airport fee is not a quote, it is a number that will be short by the airport fee every time.
What happens at the edge
No sensible system tries to handle everything. The interesting design question is what it does when it reaches the limit of what it should decide.
The rule is that it hands over rather than improvises.
An unusual request. A wedding with three vehicles and a two-hour wait between venues. It takes the details and passes them to a person, because that job needs judgment about timing you would not want a system guessing at.
A complaint. Not handled by software. It is acknowledged, the details are captured, and it goes to a person who can make a decision about a refund or a gesture.
A VIP or a corporate account with its own arrangement. Recognised, and routed to whoever handles that account.
An AI employee that never says "let me get someone for you" has been built by somebody who has not run a transport business. The handover is a feature, not a shortfall.
How long the loading takes
Honestly: it depends on your operation, and any specific number in a blog post would be invented.
A single-vehicle operator with distance pricing and one airport is a different exercise from a fifty-car fleet with a zone matrix, six vehicle classes, corporate accounts on account terms and three airports with different rules. The first is a short piece of work. The second is not.
What determines it is how many distinct pricing rules you actually run, how clean your existing rate card is, and how many procedures need writing down for the first time — that last one catches people, because plenty of operators have rules they follow consistently but have never put on paper.
A demo gives a real answer for your business, because someone can look at your rate card and tell you. That is a better use of your time than a range that would be wrong at both ends.
It runs alongside your dispatch, or on its own
Two arrangements, depending on where you are.
If you run AtoZ Dispatch, the AI employee sits on top of it. Quotes come from the same rate manager your dispatchers use, and confirmed bookings land straight on the dispatch board. One set of rates, one place they are maintained.
If you are on another dispatch system, or none, it can run on its own. The rates and procedures are loaded into it directly and it works from those.
The second arrangement is more common than you might expect, because plenty of operators want to solve the unanswered-phone problem before they think about changing dispatch software.
Where to start
If you have not read the plain-English explanation of what an AI employee actually is and how it differs from a chatbot, start there. If you already know and the question is whether it can handle your pricing, the answer needs someone looking at your rate card.
Book a demo · What is an AI employee?
Frequently asked questions
Q: How does an AI employee know my prices and how my business works?
A: It is set up with your own rate table, zones, vehicle classes, surcharges and policies. It quotes from those, not from generic industry figures.
Q: What happens if a caller asks something it was not told?
A: It hands the call to a person. The design intent is that it does not guess on anything outside what it was given.
Q: How long does it take to set up?
A: It is a conversation about how your business actually works — your rates, your rules and your exceptions — rather than a form to fill in.
Q: How do I know it is quoting correctly?
A: Test it against quotes you would have given, and review the first week of real calls. That is where any gaps appear.
Q: Can I change its answers later?
A: Yes. Rates, zones and policies change, and the material it works from is updated when they do.
Available next action: Create a downloadable DOCX file here in this chat containing the editable prose above
