There’s a list sitting in your CRM right now of people who called, asked about your services, and never signed. Two hundred of them, maybe two thousand. They’re marked “closed lost” or nothing at all, and everyone’s agreed by silent consensus that they’re dead. They’re not dead. Circumstances changed for a good portion of them, and the only reason they didn’t come back is that nobody asked. AI makes asking cheap enough to be worth doing.
Key Takeaways
- Old leads convert at a lower rate than fresh ones, but they cost nothing to acquire, which usually makes reactivation the cheapest appointments you’ll book all quarter.
- Segment before you send. Blasting the whole database is how firms get spam complaints and burn a list they can only burn once.
- AI handles the personalization and the back-and-forth replies, which is exactly the work that made manual reactivation impractical.
- Text outperforms email for this, but consent rules are stricter, so check what you actually have on file.
- Route anything ambiguous to a human fast. The AI’s job is to restart the conversation, not to close it.
Why old leads are worth more than they look
Someone contacted your firm eighteen months ago about a will. They were shopping, they got distracted, life kept moving. Since then they’ve had a child, bought a house, or watched a relative die intestate. The thing that stopped them from hiring you probably wasn’t you.
The economics are what make this interesting. A fresh lead from paid search might cost a professional services firm anywhere from $80 to $400 depending on the vertical. An old lead costs you the sending fees, which round to nothing. Even at a 2% reactivation rate, a list of 1,500 dormant contacts produces around 30 conversations. For most law firms, medical practices, and accounting firms, that’s a meaningful month.
Segment first, or don’t bother
The failure mode is obvious once you’ve seen it happen: someone exports the entire contact list, writes one generic “we’d love to reconnect” message, and sends it to everybody including current clients, vendors, and the guy who asked for directions in 2019. Complaints follow, deliverability drops, and the list is effectively ruined.
Split the database into groups that deserve different messages.
- Consultations booked but never attended, which is the warmest group by a wide margin
- Inquiries that got a quote and went quiet
- Former clients whose matter closed and who haven’t been back
- Contacts who explicitly said no, which you approach carefully or skip
- Anything older than three years with no engagement, which may not be worth the deliverability risk
Former clients are the group most firms undervalue. A medical practice with patients who stopped coming after a provider left, or an accounting firm whose client moved to software and then found out software doesn’t handle an IRS notice, both have obvious reasons to reach out that don’t sound like a sales pitch.
What AI actually contributes
Reactivation isn’t a new idea. It’s just been impractical at any real scale, because doing it well meant a person reading each record, remembering what the conversation was about, and writing something that didn’t sound like a template. Nobody has time for that across 1,500 contacts.
Personalization from the record
Given a CRM record with the original inquiry, the service discussed, and the date, a language model can produce a message that references the specifics without the awkward mail-merge feel. “You reached out last spring about setting up a trust” lands very differently from “We wanted to check in.”
Handling the replies
This is the part that matters most. Send 1,500 messages and you’ll get replies at all hours, most of them short and ambiguous. “Maybe.” “What do you charge now?” “Not right now, ask me in the fall.” A person has to answer each one within minutes or the moment passes. AI answers instantly, asks the qualifying question, and offers a time from your actual calendar. The follow-up in the fall gets scheduled automatically rather than forgotten.
Scoring who’s worth calling
Not every reply deserves the same response. Reactivation systems can rank responses by intent so your intake team spends its morning on the eight people who are ready and lets the automation nurture the rest. That prioritization is worth as much as the outreach itself.
The consent question
Text messaging gets read within minutes, which is why it works. It’s also the channel where the rules bite hardest. Under the TCPA, marketing texts require prior express written consent, and “they gave us their number when they inquired” is not automatically that. Consent can also go stale in the eyes of a regulator when years have passed.
Practical approach: check what consent language was on the form when each contact came in, and treat the ones without clear opt-in as email-only. Include opt-out instructions in every message and honor them immediately, automatically, with no exceptions. For medical practices there’s a second layer, since anything referencing a condition or treatment touches HIPAA. Keep those messages generic and route the specifics to a secure channel. If any of this is unclear for your situation, it’s a conversation worth having with counsel before the first send, not after the first complaint.
Running the campaign
Start small. Take 200 contacts from your warmest segment, send, and watch what comes back for a week. You’ll learn whether your message lands, whether the AI’s replies read naturally, and whether your intake team can absorb the volume. Scaling a campaign that’s producing bad conversations just produces more bad conversations.
Keep the opening message short. Two sentences, one question, no attachments, no pitch. The goal is a reply, not a sale. Something like: “Hi Dana, this is Chris from Harbor Law. You asked about a business formation back in March, and I wanted to check whether that’s still on your list. Happy to answer a question if it is.”
Space the sequence out. Initial message, then one follow-up four or five days later, then stop. Three attempts is the ceiling for dormant contacts. Beyond that you’re just generating complaints from people who’ve made their answer clear by not answering.
What good numbers look like
Reply rates on a well-segmented reactivation campaign typically land between 5% and 15% for text and 1% to 4% for email. Of those replies, maybe a quarter to a third turn into a booked appointment. Run that against 1,500 contacts and you’re looking at roughly 20 to 60 appointments from a list you’d already written off.
Track cost per booked appointment against your paid channels, and track opt-out rate as your safety metric. If opt-outs climb past 3% or so, the message or the segment is wrong. Stop and fix it rather than pushing through, because this is a list you can only work once every six to twelve months without wearing it out.
Where the handoff happens
Set clear rules for when AI stops and a human starts. Anything involving a fee quote, a legal or medical question, an upset former client, or a request to speak with someone should transfer immediately with the full conversation history attached. Nothing sours a reactivation faster than an automated reply to someone who just explained a complicated situation in three paragraphs.
Done right, none of this feels like automation to the person on the other end. It feels like a firm that remembered them, which, in a roundabout way, is exactly what happened.
