Give us a call today: (832) 430-3194
andwright

AI

Write the Rules Before You Turn On AI Follow-Up

The plain-language rules to write before AI sends messages for your business: its job, what it may and may not say, when a person takes over, and how it is reviewed.

™

andwright.com

AI follow-up tools can reply to new inquiries in seconds, answer common questions by text, nudge people to book, and keep conversations going after hours. That speed is valuable.

But an AI that sends messages on your behalf without clear rules can also say the wrong thing, promise what you cannot deliver, or text people who did not want to hear from you.

Before you turn on AI follow-up, write the rules. Not code. Plain-language rules that describe what the AI is for, what it may and may not say, and when a person takes over. This article walks through what to include.

01Start with the job, in one sentence

Write one sentence that describes what the AI follow-up is supposed to do. For example: "Reply to new website and missed-call inquiries within a minute, answer basic questions about our services and hours, and help the person book a free estimate."

That sentence keeps the setup focused. If someone later wants the AI to also handle complaints, collect payments, or negotiate prices, you can point back to the job and decide whether that belongs in it.

02List what it may say

An AI is only as accurate as the information it is given. Write down the facts it may share:

  • Services you offer, and ones you do not.
  • Service area and hours.
  • How booking works, and any requirements.
  • Policies such as cancellations, deposits or warranties, in your exact words.
  • Prices, only if they are fixed and you are comfortable sharing them.

Keep this in a single document that someone owns and updates. When your hours change or you stop offering a service, update the document the same day.

03List what it must not do

The "must not" list is often more important than the "may" list. Common items:

  • Do not quote prices that are not on the approved list.
  • Do not promise availability, timelines or outcomes.
  • Do not give medical, legal, financial or safety advice.
  • Do not ask for sensitive information such as card numbers, social security numbers or health details by text.
  • Do not argue with or try to persuade someone who says no.
  • Do not pretend to be a human if asked.
  • When unsure, say a team member will follow up, and flag the conversation.

04Decide when a person takes over

Good AI follow-up hands off to a person at the right moment. Write down the triggers, for example:

  • The customer asks for a person.
  • The customer is upset or describes a complaint.
  • The request is urgent, such as a leak, a lockout or a safety issue.
  • The question is outside the approved information.
  • The customer is ready to book something that needs human scheduling.

For each trigger, name who takes over and how quickly. A handoff that sends a notification to nobody is not a handoff. Who Owns the Lead? covers assigning ownership.

AI follow-up often uses text messages, and texting has rules. The AI should only text people who have given permission, identify your business, honor opt-out requests such as STOP immediately, and respect quiet hours. Texting Customers: Consent and Opt-Out Basics covers this.

If the AI makes outbound voice calls, note that the FCC ruled in February 2024 that AI-generated voices count as "artificial" under the Telephone Consumer Protection Act, which brings consent requirements with it. This is not legal advice.

06Set the tone

Describe how messages should sound: friendly, brief, plain English, no slang, no exclamation marks in every line, signed with the business name. Give two or three example messages you like. The AI will follow examples better than abstract adjectives.

Ask for short messages. A text that fills the whole screen feels like a sales script, not a conversation.

07Decide what gets logged and reviewed

Every AI conversation should be saved in your CRM, attached to the contact, so anyone on the team can see what was said. Decide who reviews conversations and how often. In the first few weeks, read a large sample daily. Look for wrong information, missed handoffs and awkward replies, then update the rules.

After it settles, a weekly review of flagged conversations and a random sample is usually enough.

08Test before customers see it

Before launch, have a few people play customers. Ask normal questions, odd questions, rude questions, and questions outside the rules. Try to get it to quote a price or promise a date. Every time it gets something wrong, fix the rules or the information and test again.

09Keep the rules short enough to maintain

The rules do not need to be long. One or two pages is usually enough for a small business. What matters is that they exist, someone owns them, and they are updated when the business changes.

AI follow-up works best as a fast, polite first response that gets people to the right next step, with a person close behind. AI Follow-Up Is Not a Chatbot Stuck on Your Website and What an AI Receptionist Can and Can't Do cover where it fits.

Want help writing the rules and setting it up?

andwright sets up AI follow-up with written boundaries, handoffs and review built in. Book a strategy call and we will start with the rules, not the tool.

Book a strategy call