On this page (9 sections)
AI lead follow-up means using software to help with the work between “someone enquired” and “they booked”: reading the enquiry, drafting a reply, reminding you to follow up and flagging leads that are waiting. Done well, it makes a small team reply faster and forget fewer people. Done badly, it sends generic messages that make you look like you didn’t read what they wrote.
My rule is simple. Let AI do the reading, sorting, drafting and reminding. Let a person read the enquiry and send anything that shapes the customer’s first impression or mentions money.
Key takeaways
- Speed matters most at the first reply. In a 2011 analysis of about 1.25 million leads, firms that tried to contact a lead within an hour were nearly 7 times as likely to qualify it (have a meaningful conversation with a decision maker) as firms that waited an hour longer (Harvard Business Review, 2011).
- AI helps you get there: it drafts while you’re busy, so a person only has to check and send.
- Keep a person on the first reply, quotes and complaints. CASL rules apply to messages software sends too.
Why follow-up is a good job for AI
Follow-up is mostly reading and remembering, and those are things software does well. Many leads that go cold don’t go cold because the business said the wrong thing. They go cold because nobody replied in time, or nobody followed up at all. I wrote about that in why good leads go cold.
It’s also where some Canadian businesses are starting, though AI use overall is still a minority. In Statistics Canada’s survey of businesses for the second quarter of 2025 (fielded April to May 2025), 12.2% reported using AI to produce goods or deliver services. Among those AI users, the survey reported text analytics at 35.7%, virtual agents or chatbots at 24.8%, and marketing automation at 23.1%, up from 15.2% a year earlier. These are 2025 survey figures for AI-using businesses, not today’s level for all businesses (Statistics Canada, 2025). Reading enquiries and drafting replies fits right into that.
What to hand to AI, and what to keep
| Job | Who does it | Why |
|---|---|---|
| Sorting enquiries by type and urgency | AI | Messy text, low risk, saves time |
| Summarising a long enquiry | AI | A person reads the summary and the original |
| Drafting the first reply | AI drafts, person sends | The draft saves typing; the person checks it fits |
| After-hours holding reply | Automation | Same message every time, instant |
| Follow-up reminders | Automation | Never forgets a date |
| Drafting follow-ups 2 to 4 | AI drafts, person approves | Low risk once the templates are good |
| Quotes and prices | Person | A wrong number is costly |
| Complaints, refunds, sensitive replies | Person | People want a human |
A simple setup
- Enquiry arrivesForm, email, social or lead ad
- Logged in one placeCRM or shared sheet
- AI reads and draftsSummary, type, suggested reply
- Person checks and sendsWithin your reply-time target
- Reminders runFollow-ups until they reply
- Get every lead into one place. If enquiries are spread across five inboxes, no tool can follow up on them. A CRM is best; a shared sheet works to start.
- Write your templates first. AI drafts better when it has good examples. Start with the ones in lead follow-up emails.
- Give the AI your facts. Services, prices you’re happy to share, service area, hours, booking link and the questions you get most. It should only use what you give it.
- Have it draft, not send, for the first reply. A person checks it against what the lead actually asked.
- Automate the reminders. A follow-up date on every lead, and an alert when a lead has waited longer than your target. See speed to lead for setting that target.
- Review weekly. Read ten drafts. Fix the template or the facts wherever the AI got something wrong.
Keep a human in the loop, and why
AI drafts can sound confident and still be wrong. They can invent a detail, promise a date nobody agreed to, or miss the one sentence in the enquiry that mattered most. A person reading before sending catches all three. Once a type of message has been right for weeks, you can decide whether it still needs checking. Quotes and complaints always do.
An AI agent that can take actions, such as booking a slot or updating your CRM, needs clear limits on what it’s allowed to do. I cover how a small team can start with that safely in AI agents for a small team.
The rules still apply
CASL covers commercial electronic messages, whoever or whatever writes them. An enquiry gives you implied consent to send commercial messages to that person for six months, and every message needs to identify you and include an unsubscribe that works (CASL, s. 6, s. 10 and s. 11). If your AI tool sends emails and a different tool sends texts, make sure an unsubscribe in one stops both. The CASL compliance checklist has the details. This isn’t legal advice.
An example, so you can picture it
This is an illustration, not a client story.
A two-person roofing company gets enquiries through its website and Facebook, often while both owners are on a roof. Every enquiry now lands in one shared list. An AI step reads each one, tags it (repair, replacement, insurance, other), and drafts a reply using the company’s own templates and service area. When one of the owners comes down for a break, they read the draft, fix anything off, and send it from their phone. Reminders handle the follow-ups. The owners still write every quote themselves.
Common mistakes
- Letting AI send the first reply unchecked. That’s where trust is made.
- Giving it no facts. Without your prices, area and hours, it guesses.
- Automating a broken process. If leads aren’t in one place, fix that first.
- Generic replies. If a draft could be sent to anyone, rewrite it so it answers what this person asked.
- Ignoring unsubscribes across tools. One unsubscribe must stop every channel.
Where this fits
AI lead follow-up is one piece of a bigger follow-up system. It doesn’t fix an unclear offer or ads that bring the wrong people, which is why I start with where growth gets stuck. If you’d like help setting this up, the Growth Diagnostic covers your follow-up, or book a call.
Sources
- Oldroyd, J. B., McElheran, K. and Elkington, D. “The Short Life of Online Sales Leads.” Harvard Business Review, March 2011. hbr.org
- Statistics Canada. “Analysis on artificial intelligence use by businesses in Canada, second quarter of 2025.” Analysis in Brief, Catalogue 11-621-M, 2025008. statcan.gc.ca
- Canada’s Anti-Spam Legislation (S.C. 2010, c. 23), sections 6, 10 and 11. Justice Laws Website
Questions people ask
Can AI follow up with leads for me?
Yes, for parts of the job. AI is good at reading an enquiry, sorting it, drafting a reply and reminding you when someone is waiting. I'd still have a person read and send the first real reply and anything involving price, because that's where trust is won or lost.
Will customers know the reply was written by AI?
They often can tell when a reply is generic or misses what they asked. That's the real risk, not the tool itself. A draft that a person checks against the actual enquiry reads like a person wrote it, because one did the important part.
Does CASL apply to AI-sent messages?
Yes. A commercial email or text is your message whether you typed it or software did. You need consent, your business details and a working unsubscribe in every one. This isn't legal advice.
What's the difference between a chatbot and AI lead follow-up?
A chatbot answers people while they're on your site. AI lead follow-up works after they've left: it reads what they sent, drafts replies, schedules reminders and flags leads that are waiting too long.