The Problem
Apex Realty was generating more than 200 inbound leads per month from their website, Realtor.ca listings, and paid ads. Their sales director was following up on every single one manually.
The problem was not volume — it was quality. Roughly 70% of those 200 leads were tire-kickers: people early in a two-year research phase, renters looking for information, or inquiries from outside their service area. The 30% who were genuinely ready to transact — pre-approved buyers, sellers with urgent timelines, investors looking to move in the next 90 days — were sitting in the same queue as everyone else.
The typical follow-up path: a lead comes in, gets added to a spreadsheet, a junior agent calls within 24–48 hours, spends 15 minutes on the phone assessing fit, and either books a showing or marks it as unqualified. With 200 leads a month, that's 50 hours of agent time spent on calls — the majority of which ended in "not a fit right now."
The deeper issue was that the high-value leads — the ones with pre-approval letters and 60-day move-in timelines — were not getting faster or better treatment than anyone else. An investor ready to buy a $900K property was waiting in the same queue as someone casually browsing condos.
Apex needed a system that could read every inbound lead, assess its quality and urgency, route it appropriately, and draft an opening message — before any human looked at it.
Our Solution
We built a multi-stage AI qualification pipeline that runs on every lead the moment it arrives. Each lead is analyzed, scored 1–10, tagged with intent and urgency signals, and automatically routed to the right follow-up path — all within two minutes of form submission.
PIPELINE ARCHITECTURE
[Website Form / Realtor.ca / Ad Landing Pages]
|
[n8n Webhook Trigger]
|
┌──────────┴──────────┐
| |
[Data Parsing] [Context Enrichment]
• Name, email • Company/employer lookup
• Message text • Location validation
• Property type • Budget range parsing
• Timeline • Source attribution
| |
└──────────┬──────────┘
|
[Claude AI Analysis]
• Intent classification
(buying / selling / investing / info-seeking)
• Urgency scoring (immediate / 0-3mo / 3-12mo / unknown)
• Budget estimation from message text
• Sentiment analysis (motivated vs. browsing)
• Red flag detection (outside service area, renter signals)
|
[Lead Scoring Engine]
Score 1–10 based on:
• Intent quality (40%)
• Timeline urgency (30%)
• Budget signals (20%)
• Message quality (10%)
|
┌───────────┼───────────┐
| | |
[Score 8–10] [Score 5–7] [Score 1–4]
Priority Standard Nurture
Route Route Route
| | |
Agent DM Lead queue Email drip
+ drafted + drafted sequence
follow-up follow-up (no agent
within 5 within 4h time spent)
minutes
|
[HubSpot CRM]
• Lead record created
• Score + tags applied
• Deal stage set
• Source tracking
• Follow-up task assignedThe AI analysis component uses Claude to read the raw lead message and form data, then returns a structured JSON assessment including:
A lead that says "we're pre-approved for $850K and need to be in before school starts in September" gets a score of 9, a priority DM to the assigned agent within 5 minutes, and a drafted response that references their timeline and budget range. A lead that says "just starting to look around for next year sometime" gets a 4, enters the nurture email sequence, and no agent time is spent until they re-engage.
HubSpot integration: Every lead creates a new contact and deal record automatically. Score, intent tag, urgency tier, and AI summary are written as custom properties. Agents see the full assessment before they pick up the phone.
Tech Stack
| Component | Tool |
|-----------|------|
| Workflow engine | n8n (self-hosted on Hetzner VPS) |
| AI analysis | Claude Haiku via API (low latency, low cost per lead) |
| CRM | HubSpot (Sales Hub Starter) |
| Lead routing | n8n conditional branches + HubSpot task assignment |
| Email sequences | HubSpot workflow automation |
| Team alerts | Slack (priority leads) |
| Enrichment | Clearbit basic lookup (company/employer) |
| Form sources | Webflow forms + Zapier connector for Realtor.ca |
Results
Lead handling before: 50+ hours/month of agent call time. 24–48 hour average response time. No differentiation between hot and cold leads.
Lead handling after (30-day snapshot):
The most direct ROI: one of those three deals in month one was a buyer who had inquired with a competing brokerage the same day. Apex responded in 6 minutes. The competitor called back the next morning. The buyer signed with Apex.
Timeline
| Day | Work |
|-----|------|
| Day 1 | Discovery call (45 min). Audited existing lead flow, HubSpot config, current response times, and pain points. |
| Day 2 | Designed pipeline architecture. Defined scoring rubric with sales director. Agreed on routing rules and agent assignment logic. |
| Day 3–4 | Built n8n workflow, Claude API integration, and JSON parsing logic. Configured HubSpot custom properties. |
| Day 5 | Built HubSpot deal creation and task assignment automation. Wrote 3 follow-up email template variants (buyer / seller / investor). |
| Day 6 | Testing with 50 real historical leads. Calibrated scoring against known outcomes (which leads actually converted). |
| Day 7 | Refinement pass based on test results. Adjusted urgency weighting after reviewing calibration data. |
| Day 8 | Agent training session (30 min Loom video). Delivered documentation. Went live. |
Total delivery: 8 days.
Investment
Core tier: $3,500
Ongoing costs to Apex: n8n VPS hosting (~$12/month), Claude API usage (~$18/month at their lead volume), HubSpot Sales Hub Starter ($20/month). Total running cost: ~$50/month for a system that recovered roughly $72,000 in commissions in its first month of operation.
The pipeline also reduced the sales director's weekly workload by ~11 hours — time she has redirected to high-value client relationship work.
*Need a qualification pipeline for your sales process? We build these for real estate, SaaS, professional services, and agencies. [Get a quote →](https://jsterlinglabs.com/contact)*