AI SMS agents — speed-to-lead, lead revival, and two-way qualification.
The UK industry average inbound response time is 47 minutes. Most CRMs are full of leads marked "lost" that just needed one message at the right moment. A bespoke AI SMS agent fires within seconds of an enquiry, revives dormant buyers at 31% conversion, and qualifies new leads over plain text — before your team touches a single queue.
45–60 min · Free · No pitch
The leads you already paid for are going cold
A buyer clicks "enquire" on a Saturday evening. They're comparing three businesses. By the time your team reads the notification on Monday morning, they've already heard back from whoever replied at the same time they hit submit. That scenario plays out every weekend, in every operation where inbound leads wait for a human.
It's not just the after-hours problem. It's the 300 leads in your CRM from six months ago that someone marked "lost" and never touched again — most of them still buyable, if the message arrived at the right moment. It's the new enquiry that needs qualifying before it's worth a salesperson's time, but qualification means a call, and half the calls don't get answered.
All three problems are the same problem: there is no AI standing between the event and the human. A bespoke SMS agent closes that gap — fires within seconds, revives the ones that went quiet, qualifies the ones that need it, and hands your team a warm lead rather than a cold queue.
What a bespoke AI SMS agent does.
Three distinct patterns — deploy one or combine them into a single pipeline.
Speed-to-lead
New enquiry arrives → personalised SMS fires within seconds. Before the buyer has checked the next listing, your response is on their phone. Personalised to the specific vehicle, service, or product they enquired about.
Lead revival
AI identifies leads that went cold 30–90 days post-enquiry and fires a re-engagement sequence personalised to their original interest. Claude generates each message from context — not from a template. Measured conversion on dormant leads: 31%.
Two-way AI conversation
Customer replies in plain English. Claude reads intent (interested, price question, time confirmation, opt-out), generates a contextually appropriate reply, and continues the conversation until it reaches a defined outcome — booking, escalation, or closure.
Structured qualification extraction
A second AI layer reads the completed conversation and extracts structured qualification data — what the lead needs, urgency, relevant constraints, pathway recommendation — into a JSON checklist in the database. Staff see a completed assessment, not a raw transcript.
Intent detection and routing
Every inbound SMS is classified: appointment request, price question, time confirmation, opt-out, maintenance reported. The classification routes the message to the right action — book a slot, escalate to human, update CRM — without anyone reading the thread first.
Proximity-optimised dispatch
When a lead qualifies for a survey or appointment, the system calculates available slots using the Haversine distance formula — clustering the booking near the engineer's existing jobs that day to minimise travel time. SMS fires automatically with the suggested slot and a one-click confirm link.
DNC compliance and quiet hours
Do-Not-Contact list checked before every send. STOP and opt-out detected and logged immediately. Configurable quiet hours (default 20:00–09:00 UK time). No message sent to a number that has opted out. Full audit log of every action.
Live UK AI SMS deployments
Both builds below use the same pattern: Twilio handles SMS delivery, Claude generates the responses, Supabase edge functions orchestrate the pipeline.
Every inbound dealer lead receives a personalised SMS within 60 seconds (vs 47-minute UK industry average — 47×). The Lead Revival Agent identifies leads dormant for 30–90 days and fires Claude-generated re-engagement sequences. Measured conversion on dormant leads: 31%. The garage variant ("Sarah") factors in MOT history, predictive maintenance, and vehicle health scores to personalise each revival message.
When a solar lead clears pre-qualification, a cron job calculates the best survey slot using the Haversine formula — clustering the appointment near the surveyor's existing jobs on that day to minimise travel. SMS fires automatically: "Hi [Name], we'd like to book your survey for [Day] at [Time]. Confirm: [link] — Different time: [link]." Customer confirms or selects an alternative slot in one click. Surveyor routes efficiently. No human needed to schedule or dispatch.
Contact form submission triggers the SMS agent "Sarah" within seconds. Sarah conducts a two-way qualification conversation — Claude Sonnet generates each reply with full context. Claude Haiku then extracts an 11-point qualification checklist into structured JSON. Staff see the completed assessment in the dashboard without reading a raw transcript. A person in need hears from the charity within seconds, not days.
The before picture.
A buyer enquires on a Saturday afternoon. The response arrives Monday morning. They booked with whoever replied first.
Your CRM has 340 "lost" leads from the last six months. Most of them would still be worth a message. Nobody has time to send 340 messages.
A new enquiry arrives. Someone has to call them to qualify. Half don't answer. A quarter call back at inconvenient times. Qualification takes a week.
A counselling charity receives a contact form submission from someone in need. The next human contact is a callback, scheduled for two days later.
The after picture.
Every enquiry receives a personalised SMS within 60 seconds — whether it arrives at 2pm on a Tuesday or 10pm on a Sunday. Your response is first, every time.
The lead revival agent reviews your dormant contacts, identifies the right re-engagement window, and fires Claude-generated sequences referencing their original interest. 31% re-engage.
New enquiries are qualified over SMS before a salesperson touches them. By the time your team picks up the lead, the AI has already established interest, timeline, and fit.
A contact form submission triggers an immediate SMS conversation. The person in need hears from the charity within seconds, not days. A structured qualification checklist sits in the dashboard before a counsellor makes contact.
How it works
An event fires the pipeline — a new lead in the CRM, a form submission, a scheduled revival check. The trigger is typically a Supabase database function (AFTER INSERT trigger or scheduled cron) rather than application-level code, so it fires regardless of whether the web app is running.
The edge function pulls the customer's context: name, what they enquired about, any conversation history, CRM data, service history, vehicle details, or qualification state. This context is passed to Claude alongside the system prompt that defines the agent's persona and rules.
Claude Haiku (for speed and volume) or Claude Sonnet (for complex qualification) writes a natural-language reply specific to this customer and this moment. It is not a template with placeholders — it is a generated message that reads like a thoughtful human wrote it.
The message is sent via Twilio Programmable SMS to the customer's phone. Status tracking (sent / delivered / failed) is logged back to the database. The customer receives a plain SMS on their existing number.
Twilio fires an inbound webhook to the edge function. The reply is classified by intent (interested, price question, time confirmation, opt-out, unclear). Claude reads the full conversation history and generates the appropriate next message, or closes the conversation and routes to a human.
A second AI layer extracts qualification data from the conversation into a structured JSON record in the database. Staff see a completed checklist rather than a raw thread. The CRM is updated with the qualification state, next action, and full transcript.
Why bespoke beats a bulk SMS tool.
A bulk SMS tool sends a templated message to a list. Every recipient gets the same text with a name placeholder.
A bespoke AI SMS agent generates each message from the recipient's actual context — their specific enquiry, their vehicle, their service history, their original interest. The message reads like it was written for them.
Why it matters: Conversion on personalised messages is materially higher than on templates. The 31% lead revival rate is built on generated messages, not blast campaigns.
A bulk tool fires once. If the lead doesn't reply, that's the end of the workflow.
A bespoke agent manages the full conversation: fires the opener, reads every reply, classifies intent, responds appropriately, and handles the conversation through to a defined outcome.
Why it matters: Most conversions happen on the second or third exchange. A tool that stops after one message leaves those conversions on the table.
A bulk SMS tool delivers messages. What the lead says back has to be read and actioned manually.
A bespoke build extracts structured data from the conversation — qualification state, intent, next action — into the CRM automatically. The human's first contact is with a pre-qualified lead, not a raw inbox.
Why it matters: Qualification is the part that takes human time. Automating it means your team's first call is to a lead that's already been assessed.
Compliance and control
SMS outreach carries regulatory obligations. Every build includes:
Any STOP, UNSUBSCRIBE, or semantic opt-out is detected immediately, logged to the DNC list, and no further messages sent. Twilio also enforces carrier-level compliance independently.
Every outbound message checks the DNC table before sending. Numbers added manually by staff or automatically by opt-out are permanently excluded from all future sends.
Configurable send window (default: no outbound messages between 20:00 and 09:00 UK time). Scheduled sends queue until the window opens. Revival messages respect the same constraint.
Every sent and received message is logged with timestamp, direction, Twilio SID, and delivery status. Every AI-generated reply is stored alongside the conversation context that produced it. Full audit trail for any conversation.
Common questions about AI SMS agents
An AI SMS agent is a bespoke system that sends, receives, and manages text message conversations on behalf of your business — using Claude to generate contextually aware replies rather than templates. The result is a lead pipeline that operates 24/7 without a human in the loop until the lead is qualified and warm.
Still have a question? Book a discovery call — direct line to me, Dean.
Where this is running. What to read next.
WheelsAI — speed-to-lead and lead revival for UK car dealers
60-second SMS response, 47× faster than industry average, 31% conversion on dormant leads. Full DMS and garage revival system.
Read the case studyCASE STUDYCounselling in the Community — two-way SMS qualification agent
Contact form to Claude-driven SMS conversation to 11-point structured qualification checklist — for a King's Award counselling charity.
Read the case studyGUIDEVoice Agents
The phone-call equivalent — inbound qualification, appointment booking, outbound follow-up over voice rather than text.
Read the guideEvery week cold leads sit untouched, the revival window closes.
The 30–90 day dormancy window is when re-engagement works. After that, conversion drops sharply. Meanwhile, every enquiry arriving outside business hours is being decided before anyone reads it on Monday.
A 45–60 minute discovery call maps your lead database, your typical cold-lead volume, and what a 31% revival rate would mean in recovered revenue. You walk away with a costed build map regardless.
Map my lead response gap