Charge for the pilot. A free live deployment costs you wholesale fees and minutes, and the data we found points to paid pilots closing better. Here is the 30-day structure, the math, and the two times free is fine.
Lumina7 min read
Charge for it. A free pilot of a live AI receptionist costs you real minutes and real setup money, and it teaches the owner that the thing is worth nothing. Run a short paid pilot instead, with a written finish line and the price of the full deal stated on day one. If the owner flatly refuses to pay anything, the free demo is your free trial, and a second free one is not.
That is my view. Here is the reasoning, the numbers, and the two cases where I would break my own rule.
Do free pilots convert for AI services?
Mostly they don't, and the people who track this say so in plain terms. A Rework guide on pilot programs describes an unnamed enterprise software company that watched pilots for 18 months. Pilots with written criteria, a fixed timeline, a fee and pre-agreed conversion terms closed at 68%. Free pilots with fuzzy goals closed at 23%. Treat that as one company's anecdote, because the guide names no source and the deals were much bigger than yours. The direction still matches what you will see at the kitchen table of a roofing contractor.
Jason Lemkin at SaaStr says paid pilots convert "from about 60% to over 90%" depending on how much deployment work sits in the way. His piece says nothing about free pilots, so it can't tell you what free costs you. It does tell you a paid pilot is a normal, survivable thing to ask for.
Why would free convert worse? A Stackmatix write-up on pilot conversion says free pilots "almost never die from a bad product." They die of open-ended terms, no one with a checkbook in the room, and a price that first appears after weeks of free value. The champion likes the demo. The person who signs checks never heard of it.
What does a free pilot cost you?
More than zero, and the cost is on your side.
Picture the top tier build. On the pricing section that is $1,150 to $1,400 setup and $315 to $460 a month at wholesale, plus $0.35 a minute on voice. A pilot that runs on the owner's real phone line for 30 days is not a demo. It is a live deployment, and you owe wholesale on it whether or not they sign.
Take an invented but ordinary month: 380 minutes of calls. That is $133 in voice charges alone. Add the first month's wholesale at the low end, $315, and you are $448 in before counting setup, before counting your own hours on intake, and before the owner says "let's think about it." Do that with five prospects and you have bought a very expensive lesson.
The free demo is different. It runs against a sandbox, the owner calls it, and the 100 live voice minutes a month on the $49 plan cover roughly forty demo calls. Nothing in it touches their phone line. Give that away to everyone. Giving away the live line is where the money goes.
What should a paid pilot look like?
Short, written, and priced.
HubSpot's sales blog says not to give a pilot away and to charge a nominal fee when setup takes your time, so the contact has a budget line to point at when the full contract comes up. It also says to start by asking what success looks like and work backward. I agree with both and would add numbers from the trade.
Here is the shape I would use. Thirty days, with the end date written into the agreement before anything gets built. A fee that covers your wholesale for the month plus your hours, and for a Tier One build I would not go below the wholesale setup fee. One success number the owner picks, like missed calls answered after 5pm or bookings from after-hours calls, whatever keeps them up at night.
Put the full price, setup and monthly, on page one, with a line saying the pilot fee comes off setup if they keep going.
Then book a checkpoint on day 14 and look at the calls together.
Rework's guide suggests setting the target a bit under what you normally deliver, so a pilot hits its number. If the agent usually books six after-hours jobs in a month, put the line at four.
Who has to be in the room
The person who signs. Not the office manager who loved the demo and told you so twice. No owner on the kickoff call? Then the pilot is a favor to your champion. Day 31 comes and nobody can say yes.
When is a free pilot actually fine?
Twice, in my book.
First, a trade reference you need. Your first client in a given trade, someone with a visible reputation, who will let you use their name and take a call from the next prospect. You are swapping build costs for a case study. Call it that, in your own notes. If you have no references yet, the no-case-studies post covers what to do instead of burning money on one.
Second, a tiny scope. An after-hours text-back on a single number, with a hard cap on minutes, costs you almost nothing. I would still put an end date on it.
Outside those two, I get annoyed when people call it a pilot. Unpaid work, nicer name.
What if the owner says no to paying anything?
Then the objection is usually something else. Maybe they don't believe the numbers, or they have been burned by a marketing agency, or they don't answer their own phone and would rather not know what they are missing. Money is the easy thing to say.
Go back to the demo. Put their ticket price and close rate into the ROI figure, run the audit on their real domain, and let them hear it handle the call they dread. The demo call playbook has the sequence I would follow. If they still want a discount on the deal itself, trade terms instead of cutting the number, as laid out in handling discount requests.
A flat no after all that is information. Move on. A shop that will not spend $400 to test something is not going to spend $400 a month to keep it.
Who this advice does not fit
Two kinds of seller.
If you are selling to multi-location groups with a procurement process, a free or heavily subsidised pilot is common and sometimes the price of admission. The numbers above came mostly from software sold to bigger companies, and I have not run a pilot at that scale. Your rules will differ.
And if you have no margin room, meaning your retail is close to your wholesale, any pilot is a loss. Fix the price first. The margin math post walks through it.
FAQ
How long should an AI receptionist pilot run?
Thirty days works for most local businesses. Stackmatix suggests 30 to 60 days for software pilots in general, and it says a 90-day pilot tends to lose urgency, though it offers no data for that. A month covers a full billing cycle and enough calls to judge.
Should the pilot fee come off the final price?
Yes. To the owner it looks like a deposit, not a cost. Day 30 gets a deadline. Write it into the pilot agreement so nobody argues about it later.
Can I use the free demo as the pilot?
Use it as the trial. It runs against a sandbox, so bookings are simulated and nothing hits their calendar. That is fine for proof of concept and not enough to prove it works on their line.
What do I do when a pilot goes quiet?
Hold the day 14 checkpoint no matter what. If the owner skips it, they have told you where the deal stands. Send the call data and a dated follow-up tied to when they next plan spending.
Your next step
Pick the one prospect you are most tempted to give a free month to. Write the 30-day agreement for them today, with one success number and the full price on page one, and send it before you build anything. If you want the demo side set up first, open the live demo or start on the platform and send that link as the free trial.
