AI should text your new leads only after they write back. Everything sent before that should be written in advance, by a person, and checked.

That runs against how AI texting usually gets sold. The pitch is software that answers every lead, writes every message and books appointments while you sleep. Part of that pitch is right. Speed matters, and a system should answer your leads, not your memory.

The part that is wrong is "writes every message". The texts sent before a lead speaks carry the most risk and need the least creativity. Those should be fixed in advance.

There is a second half to our stance. An AI setter, software that texts leads to get them booked, cannot fix a lead supply problem. If you do not have enough leads, AI just talks to the few you have a little faster.

What most owners believe about AI texting

Most local owners are busy doing the actual work. They are on a roof, in a chair with a client, or behind a camera. They will not call every lead in five minutes. They will not follow up seven times. Nobody can while they are doing the job.

So the belief goes like this. Turn on an AI, let it write and send everything, and bookings go up.

The speed part of that belief is correct. The Lead Response Management study by Dr. James Oldroyd with InsideSales found the odds of contacting a lead drop 100x between 5 minutes and 30 minutes. The odds of qualifying that lead drop 21x. Our own alarm ladder for a new lead runs at 5, 15 and 60 minutes. An alarm set later than about 15 minutes only measures a drop that already happened.

So yes, something has to answer fast. The question is what that something should be allowed to write.

Why letting AI write everything costs you

The first texts a lead gets are the ones with rules attached. They carry consent wording, opt-out wording and facts about your business. Those need to be exact every time. A tool that writes fresh words for every lead can drift from them.

Three examples of where that drift hurts:

  • Consent. Texts only go to people who gave written consent. A cold list is a call list, not a text list. The consent wording has to match what the person agreed to.
  • The word "cancel". We never invite the word "cancel" in a text. CANCEL is also a carrier opt-out word. We have seen a person text "Cancel" about one appointment reminder and get cut off from texts for good.
  • Facts about your business. Your hours, your service area and what you include are known before any chat starts. If software improvises them, the lead's first picture of your business can be wrong.

None of those messages needs creativity. They need to be right. Asking AI to write them adds risk and adds nothing.

The bigger cost is buying AI to solve a problem you do not have. We come back to that below.

What we write in advance

Our rule is simple. Pre-write anything sent before the lead speaks. That means the opener, the consent and opt-out wording, the final attempt, and any fact about the business.

Here is the test we use. If a message is fully decided by facts known before the chat, it is a template. Write it once, check it, and send it the same way every time.

That covers more than you might think. A confirmation text is a template. So are the reminders we send 24 hours, 3 hours, 30 minutes and 5 minutes before an appointment. So is the chase for people who did not book, the recovery after a no-show, and the long-term nurture. Our standard sequences send at least 10 texts and 10 emails in days 0 to 7. Every one of those is written in advance.

Pre-written does not mean stiff. Each touch asks one easy question in your voice. For example, ask whether the job is for their home or their business. The booking link is the reward for a reply. We do not stuff it into every message. We cover the math of how many touches it takes in follow-up economics.

What we let AI answer

Once the lead writes back, the conversation changes. Now the next message depends on what they said. You cannot pre-write every possible answer. This is where AI earns its keep.

A reply pauses the automated sequence and goes to a person. From there, AI can help that person answer fast, inside limits we set:

  • It only uses facts we gave it. No guessing about price, timing or what is included.
  • It asks one question at a time, the same way the templates do.
  • It sends the booking link once the lead has answered, as the reward.
  • It hands the chat to a person the moment it hits something outside its limits.

In our accounts, setters qualify, book and follow up by email, text and voicemail. AI helps them reply faster. It does not replace the person who owns the conversation.

That split keeps the risky messages fixed and puts the AI where it is useful: in the replies that cannot be written ahead of time. We wrote more broadly about where AI fits in a funnel in AI marketing for high-ticket services.

Why an AI setter cannot fix a lead supply problem

Here is a scenario we use with owners. Say we deliver 40 leads. The owner calls six of them once, books two, and decides the ads do not work. In that case, follow-up really is the problem. A system that answers every lead in minutes, and keeps asking easy questions, would help.

But that only works when the leads exist. When we planned an AI setter, both candidate businesses sat under our own kill line of 5 leads a month. So the plan called for 14 days of booking by hand first.

The logic is simple. If a manual chase does not move bookings, follow-up is not the bottleneck. An AI that chases faster will not move them either. The money belongs upstream, in the ads and the offer.

This is why the thing worth paying for is a kept appointment, not a lead. Leads only matter if someone answers them, and answering only matters if there are enough leads to answer. For how we connect the two for local owners, see how we run marketing for local service businesses.

Where this stance has limits

Our stance is not "AI bad, templates good". It has edges.

First, AI in the replies still needs a person watching. Limits only work if someone checks the chats and tightens the rules when the AI drifts.

Second, templates go stale. When your hours, prices or service area change, the pre-written texts and the facts the AI uses both need an update. Pre-written means someone owns the writing.

Third, a flood of leads is its own problem. If your team cannot handle a flood, we start ads at $50 to $75 a day. Cheap leads move the bottleneck to people. AI can speed up replies, but a person still has to show up and do the job.

Fourth, not every owner wants the same pace. We go lighter when an owner asks for a gentler touch. The rule about pre-writing stays the same at any pace.

What to do this week

You do not need new software to act on this. Start here:

  1. Count your leads. Look at last month. If you are under 5 a month, fix lead supply before you buy an AI setter.
  2. Chase by hand for 14 days. Answer every lead fast and follow up yourself or with your team. Watch whether bookings move.
  3. Write your templates. The opener, the consent and opt-out wording, the final attempt and a one-page fact sheet about your business.
  4. Search your texts for "cancel". Take it out of every reminder and confirmation.
  5. Set the reply rule. A reply pauses the automated texts and goes to a person, with AI helping inside clear limits.

FAQ

Can AI text new leads for a local business?

Yes, but only after the lead writes back. The first text, the consent wording, the opt-out wording, the last attempt and facts about your business should be written in advance. AI is useful for the replies that cannot be written ahead of time.

What should the first text to a new lead say?

It should be pre-written and ask one easy question, in your voice. Keep the booking link as the reward for a reply. Never invite the word "cancel", because it is also a carrier opt-out word.

What texting rules apply when AI texts leads?

The same rules as any business text. Texts only go to people who gave written consent, and the consent wording must be exact. That is why we keep those messages pre-written and checked.

Will an AI setter get me more bookings?

Only if follow-up is your bottleneck. If you get fewer than 5 leads a month, an AI setter has almost nothing to work with. Chase your leads by hand for 14 days first. If bookings do not move, put the money into the ads and the offer instead.

If you want help deciding what to automate and what to keep personal, you can start with a discovery call with our team.