A video about installing an AI bot in a barbershop drew several of its most-liked comments over one phone call. In "Barber Talks w/ @al_clipone; Installing GrokBot" on the Dan Martell channel, the bot calls customers after their haircut to ask how it went and whether friends or family need a cut. "This is how you lose business overnight lul" (20 likes), and: "If I get a haircut and a robot calls me back to ask me how it was. And if I have any family members that need to cut, that would be the last time I go to that Barber." (19 likes). The commenters shared one instinct: the call that saves the shop time may cost it customers.
Why the follow-up call lands badly
The call that upset the commenters does a job for the shop, not for the customer. The customer answers a call they didn't expect, gives attention to a machine, and gets asked for something that helps the shop: names of friends and family. A referral ask is the customer doing the shop's marketing. The same question from the barber, in person, is a different transaction: a favor between people who know each other. The bot version keeps the ask and removes the personal context that made it appropriate.
The line: who is the contact for?
Two questions sort any automated customer contact: did the customer expect it, and who does it help?
Contact customers expect, and that helps them:
- Booking confirmations and appointment reminders, tied to something they bought
- Order, shipping, and delivery updates
- Receipts and account notices
- Replies to a question they sent first
Contact that serves the business:
- Review and referral requests
- Win-back offers and "we miss you" messages
- Post-visit follow-ups with an upsell attached
- Calls that ask for names, like the barbershop referral call
Expected messages are usually easier to accept when they are timely, accurate, and tied to something the customer started: the customer wanted the message, and a machine producing it changes little. The second list asks the customer to do work for the business: a favor, judged by the person asking. A person can weigh whether the request suits the customer and take no for an answer; a script asking on a schedule can't. The barbershop comments show what that mismatch can produce.
Why phone calls need extra care
A call interrupts the customer's day in a way a text or email doesn't, and it's also the channel with legal limits. In the US, the FCC ruled on February 8, 2024 that calls made with AI-generated voices are "artificial" under the Telephone Consumer Protection Act, the federal robocall law, so its consent rules apply to them. Consent requirements depend on the type of call and its context, and unsolicited automated calls can break the law. A referral call from a shop bot, placed to customers who never agreed to automated calls, is the kind of call those rules exist for. Check current FCC guidance or ask a lawyer before setting up automated calls; this post is literacy, not legal advice.
The middle option: AI drafts, a person sends
The business-serving contacts, review asks and follow-ups among them, can still use AI: it writes the draft, the person reviews it, and it goes out under the person's name. The judgment about who gets asked, and when, stays with the person whose name the customer knows. How to Use AI for Email Without Sounding Like AI covers the draft brief: what happened, who's reading, what should happen next, how you'd say it. Human review takes more time, and it adds a decision the automation would skip: whether the message is right for this customer this week.
FAQ
Is it legal for AI to call my customers?
In the US, an AI-generated voice call can fall under the Telephone Consumer Protection Act's rules for "artificial" voices, per the FCC's February 2024 ruling. Consent requirements depend on the type of call and its context, and unsolicited automated calls can break the law. Check current FCC guidance and get legal advice for your situation before setting up an AI caller.
Do I have to tell customers they're talking to a bot?
Some places require it in specific cases. California's bot-disclosure law, in effect since July 2019, makes it unlawful to use a bot online with the intent to mislead someone about its artificial identity in order to drive a sale or influence a vote; the law provides a safe harbor when the bot clearly and conspicuously discloses that it's a bot. Disclosure can help trust as well: customers may notice when they're talking to a machine, and the undisclosed part becomes the story they tell.
Which customer messages can I automate?
The expected ones: booking confirmations, appointment reminders, order and delivery updates, receipts, and replies to a question the customer sent first. Expected contact can still land badly when it's wrong or hard to escape, so keep the messages accurate and keep an easy way to reach a person in every automated message.
Should AI answer incoming customer questions?
Yes: incoming questions are the customer asking for help, the contact they expect. Automating them can save time without much trust risk when the answers are accurate and a person is easy to reach. How to Use AI for Small Business Operations walks through the setup, including a kickoff prompt for drafting answers customers can correct.
Want to go further?
Two posts pair with this one: How to Use AI for Small Business Operations covers the incoming side of customer contact, and How to Use AI for Email Without Sounding Like AI covers the draft-and-review setup for messages that go out under your name. Sort this week's outbound contacts into the two lists above, then decide where human review belongs.