1. System Architecture
Do not let three AIs independently contact prospects. One CRM. One lead record. One decision path.
LEAD SOURCE
↓
ISBS Lead Leak / Website Leak / Follow-Up Failure / Referral / Outbound
↓
GOHIGHLEVEL — System of Record
↓
AI ORCHESTRATION
├── ChatGPT: score + diagnose + next best action + messaging
├── Gemini: Gmail / Drive / Calendar relationship intelligence
└── Claude: A-level account research + deal red-team
↓
GHL WRITES BACK
↓
James receives: WHO to contact + WHY + WHAT to say + WHAT to sell + WHEN
↓
Human conversation / approved outreach
↓
Transcript + outcome returns to GHL
↓
AI updates score, objection, offer, stage, next action
Security rule: Never place OpenAI, Anthropic, Google, or HighLevel API keys inside this HTML/WordPress widget. Keys belong in a secure server-side integration layer, automation platform, or protected environment variable.
0
Freeze the Revenue Process
Adopt one sales promise
Use: “We find where leads are leaking, fix the failure point, and install a system that keeps opportunities from disappearing.” Put this language in your internal SOP and sales scripts.
Adopt the 5-step offer ladder
1) Free assessment → 2) diagnosis/audit → 3) prescription → 4) implementation → 5) optimization. Stop creating unrelated offers until this ladder is selling.
Define AI autonomy
Phase 1 rule: AI may research, score, tag, draft, summarize, and recommend. James approves personal prospect messages, pricing, proposals, and negotiations.
Choose a pilot cohort
Use 10–25 actual ISBS leads. Include a few assessment completers, dormant contacts, and new prospects. Do not start with the entire database.
1
Build the GHL Revenue Intelligence Layer
Contact Custom Fields — folder: ISBS Revenue Intelligence
| Field | Type | Purpose |
|---|---|---|
| ISBS ICP Fit Score | Number | 0–100 fit to ideal client |
| ISBS Lead Intent Score | Number | 0–100 behavioral intent |
| ISBS Primary Leak | Dropdown | Follow-Up / Website / Capture / Routing / Nurture / Unknown |
| ISBS Secondary Leak | Dropdown | Same options |
| ISBS Leak Evidence | Long Text | Observable evidence; distinguish facts from inference |
| ISBS Relationship Strength | Number | 0–100 existing relationship strength |
| ISBS Buying Signals | Long Text | Assessment activity, replies, booking behavior |
| ISBS Last AI Review | Date | Last intelligence refresh |
| ISBS AI Confidence | Number | 0–100 confidence in recommendation |
| ISBS AI Summary | Long Text | One-paragraph current account summary |
Opportunity Custom Fields — folder: ISBS Deal Intelligence
| Field | Type | Purpose |
|---|---|---|
| Deal Priority Score | Number | 0–100 commercial priority |
| Revenue Potential | Dropdown | Low / Medium / High / Strategic |
| Primary Objection | Dropdown | Price / Timing / Trust / Existing CRM / Need / Authority / Other |
| Recommended Offer | Dropdown | Audit / Lead Recovery / Website Conversion / Follow-Up / Custom |
| Next Best Action | Long Text | Specific action, not “follow up” |
| Next Action Date | Date | Date action should occur |
| Action Owner | Dropdown | James / GHL / AI / Team |
| Deal Risk | Dropdown | Low / Medium / High |
| AI Deal Thesis | Long Text | Why this prospect should buy |
| AI Deal Red-Team | Long Text | Strongest case against buying |
Create the Contact custom-field folder and all 10 fields above
In GHL: Settings → Custom Fields → Contact → create folder → create fields.
Create the Opportunity custom-field folder and all 10 fields above
Use Opportunity fields for deal-specific intelligence. Do not accidentally create them as Contact fields; HighLevel does not let you switch the object type later.
Create pipeline: ISBS Revenue Pipeline
Stages: New → Assessment Complete → AI Qualified → Contact Now → Engaged → Discovery Booked → Audit/Diagnosis → Proposal → Decision → Won. Add Nurture and Lost as terminal paths.
Set pipeline card visibility
At minimum show Opportunity Value, owner, next-action date, and priority where your GHL layout allows it.
2
Create the Lead Scoring Standard
ISBS PRIORITY SCORE =
30% ICP FIT
25% LEAK SEVERITY
20% INTENT
15% ECONOMIC VALUE
10% RELATIONSHIP / ACCESS
A = 80–100 → Human contact same business day
B = 65–79 → Personalized nurture + research
C = 45–64 → Automated nurture / monitor
D = 0–44 → Do not spend human selling time
Override:
Any explicit request for help, pricing, audit, booking, or implementation = A until reviewed.
Document the scoring rubric in GHL notes or your internal SOP
Prevent the AI from inventing a different definition every time.
Test scoring against 10 known leads
Ask: “Would I personally call these A leads today?” If not, fix the rubric before automation.
3
Build the First GHL Trigger: Assessment → Revenue Agent
Workflow: ISBS — Assessment Revenue Qualification
TRIGGER
Assessment form / survey submitted
ACTIONS
1. Create or update Contact
2. Set source = Lead Leak Assessment
3. Create Opportunity → stage: Assessment Complete
4. Write assessment scores / leak category
5. Send data to secure AI endpoint via Custom Webhook
6. AI returns structured intelligence
7. Write returned values to Contact / Opportunity fields
8. If Priority ≥ 80 → move to Contact Now
9. Create task for James
10. Send internal notification containing:
• why this lead matters
• likely leak
• next best action
• 30-second call opener
11. If Priority 65–79 → move to AI Qualified and start personalized nurture
12. Else → Nurture
Minimum outbound webhook payload
{
"contact_id": "{{contact.id}}",
"first_name": "{{contact.first_name}}",
"last_name": "{{contact.last_name}}",
"email": "{{contact.email}}",
"phone": "{{contact.phone}}",
"company": "{{contact.company_name}}",
"website": "{{contact.website}}",
"source": "Lead Leak Assessment",
"assessment_total": "{{custom.assessment_total}}",
"follow_up_score": "{{custom.follow_up_score}}",
"website_score": "{{custom.website_score}}",
"capture_score": "{{custom.capture_score}}",
"submitted_at": "{{date.now}}"
}
Replace placeholder merge-field names with the actual GHL merge fields in your sub-account.
Create workflow and trigger
Use the actual Lead Leak form/survey submission trigger.
Create/update the opportunity automatically
Place it in Assessment Complete and prevent unwanted duplicate opportunities.
Create secure AI endpoint
Use a protected server, cloud function, Make/Zapier/Pipedream-style integration, or other server-side layer. Do not expose credentials in this page.
Configure GHL Custom Webhook to send lead data
POST JSON. Include only fields needed for qualification. Add authentication at the endpoint.
Force structured AI output
Require JSON containing scores, leak diagnosis, evidence, next best action, offer, confidence, and a call opener.
Write AI results back to GHL
Map values into the fields from Phase 1. No useful intelligence should live only in a chat transcript.
4
Install the ChatGPT Revenue Chief Prompt
SYSTEM ROLE: ISBS Revenue Chief
Your job is not to create activity. Your job is to create qualified sales movement.
For every lead:
1. Separate known facts from inference.
2. Score ICP fit, leak severity, intent, economic value, and relationship/access.
3. Calculate a 0–100 Deal Priority Score.
4. Identify the Primary Leak and supporting evidence.
5. Identify the strongest buying signal and strongest risk.
6. Recommend exactly ONE Next Best Action.
7. Name the Action Owner: James, GHL, AI, or Team.
8. Recommend the narrowest ISBS offer that solves the diagnosed problem.
9. If evidence is weak, lower confidence rather than inventing certainty.
10. Do not recommend a full CRM replacement unless the evidence demands it.
11. Human contact outranks automation for A leads.
12. Output valid structured JSON for GHL plus a short human-readable brief.
Required output:
priority_score
icp_fit_score
intent_score
revenue_potential
primary_leak
secondary_leak
leak_evidence
buying_signals
primary_objection
recommended_offer
next_best_action
action_owner
deal_risk
ai_confidence
ai_summary
call_opener
Save this prompt as the canonical ISBS Revenue Chief prompt
Use the same version in every automated qualification call. Version it when changed.
Add a “facts vs inference” requirement
No prospect should be approached based on an invented complaint or unsupported claim.
5
Build the Daily Revenue Brief
DAILY ISBS REVENUE BRIEF
1. MONEY NOW
A leads requiring human action today
2. OVERDUE
Leads whose Next Action Date has passed
3. NEW QUALIFIED LEADS
New A and B opportunities since last review
4. STALLED
Engaged opportunities with no meaningful movement
5. TODAY'S CONVERSATIONS
Meetings/calls + 5-point prep brief
6. REPLIES THAT MATTER
Prospect responses requiring judgment
7. PIPELINE
Opportunity count/value by stage
8. PATTERN INTELLIGENCE
Repeated objection, leak, or buying signal
9. JAMES'S THREE MOVES
Exactly 3 highest-value human actions
Create GHL Smart Lists / filters for A leads, overdue actions, and stalled deals
Build the datasets before trying to summarize them.
Create a daily export or secure API pull
First version can use a Google Sheet bridge. Mature version should query the CRM directly through a secure integration.
Generate the brief using the exact 9-part structure above
The system is failing if the output is a generic “here is your pipeline” summary.
6
Give Gemini a Narrow Job: Google Relationship Intelligence
GEMINI ASSIGNMENT — RELATIONSHIP INTELLIGENCE
For this company/contact, search the authorized Google information available to you.
Return:
• prior email exchanges and most important dates
• commitments made by either side
• prior meetings
• Drive files/proposals connected to the account
• explicit needs, objections, timing, or budget clues
• unresolved follow-ups
• relationship-strength evidence
• contradictions or changes in the prospect's position
Do not provide a sales strategy.
Do not invent facts.
Separate direct evidence from inference.
Return a concise intelligence packet for the ISBS Revenue Chief.
Connect only the Google services you are comfortable authorizing
Start with Gmail/Drive/Calendar only when useful. Apply least privilege.
Use Gemini only for B/A leads or accounts with prior history
Do not burn time mining Google history for cold C/D prospects.
Write useful relationship findings back to GHL
Update Relationship Strength, Buying Signals, AI Summary, or notes.
7
Give Claude a Narrow Job: A-Lead Strategy + Red Team
CLAUDE ASSIGNMENT — DEAL RED TEAM
You are not the salesperson. You are the skeptical buyer/CFO.
Inputs:
• company/account research
• assessment data
• relationship intelligence
• call transcript / notes
• current proposed ISBS solution
• proposed price if available
Return:
1. Strongest case that the prospect should NOT buy.
2. Evidence missing from our current thesis.
3. What we may be overdiagnosing.
4. What we may be underselling.
5. Most likely objection.
6. Minimum solution that could solve the actual problem.
7. Proof/ROI evidence required before proposal.
8. One critical question James must ask before advancing.
9. Deal risk: Low / Medium / High.
10. Recommendation: ADVANCE / HOLD / NURTURE / DISQUALIFY.
Do not flatter the sales team.
Do not rewrite generic marketing copy.
Challenge assumptions.
Trigger Claude only for priority A or strategic opportunities
Use the second model where disagreement is valuable, not everywhere.
Store the red-team conclusion in GHL
Write to AI Deal Red-Team, Deal Risk, and the opportunity note.
Require resolution before proposal
James must answer the critical missing question or explicitly accept the risk before Proposal stage.
8
Install the Call → Intelligence → Follow-Up Loop
AFTER EVERY SALES CALL
INPUT:
Transcript OR notes
AI EXTRACTS:
• stated pain
• quantified pain
• current process
• technology already in place
• decision maker(s)
• urgency / timeline
• economic clues
• buying signals
• objection
• commitments
• next step
• recommended pipeline stage
• recommended offer
• follow-up message
WRITE BACK:
GHL note + custom fields + stage + next-action date
THEN:
James approves external follow-up for A leads.
Choose one capture method for sales calls
Transcript, meeting notes, or standardized post-call form. “I will remember it” is not a system.
Create post-call extraction prompt / automation
Use the exact fields above so every deal receives consistent intelligence.
Require a Next Action Date before leaving Engaged or later stages
A deal without a next date is a stalled deal, regardless of how positive the call felt.
9
Build Controlled Nurture
Recommended lead treatment
| Tier | Treatment |
|---|---|
| A | Human contact; personalized follow-up; meeting prep; red-team before important proposal. |
| B | Personalized message + proof/case content + specific CTA. Re-score on engagement. |
| C | Automated education around their leak category. Re-score on click/reply/assessment completion. |
| D | Minimal nurture or suppress. Protect human selling time. |
Build separate nurture paths by Primary Leak
Follow-Up Failure recipients should not receive the same message as Website Leak recipients.
Add re-score events
Reply, appointment booked, repeat assessment, important page visit, or direct request should trigger review/escalation.
Create an A-lead human-contact guardrail
Do not let a strong prospect receive six robotic emails because the workflow was easier than making the call.
10
Test Before Launch
Submit a fake Lead Leak Assessment
Verify contact creation/update, opportunity creation, stage, assessment fields, and source.
Verify outbound webhook payload
No blank required fields. No sensitive credentials. Correct contact identifier.
Verify AI JSON
Valid structure, reasonable scores, facts/inference separation, one next action.
Verify write-back to GHL
Fields and opportunity stage update correctly; no duplicate opportunities.
Verify A-lead escalation
James receives a usable internal notification and task, not an unreadable JSON blob.
Verify B/C nurture routing
Correct category, correct message, correct delay, correct exit conditions.
Run the system on the pilot cohort
Compare AI priority against your own judgment. Investigate large disagreements before scaling.
11
Launch Metrics + Weekly Improvement
Track speed-to-first-human-touch for A leads
Goal: know the median and reduce it.
Track Assessment → Conversation conversion
This exposes whether the diagnostic is actually starting sales conversations.
Track Conversation → Audit/Diagnosis conversion
This tells you whether the opening conversation is compelling.
Track Proposal → Won rate and average sales cycle
Use the red-team data to understand lost deals.
Run a weekly Revenue Council review
Top wins, top losses, scoring mistakes, recurring objections, content opportunity, workflow failure, next experiment.
Definition of Done — Version 1
V1 is live only when: a test assessment automatically creates/updates the GHL record, receives an AI qualification, writes intelligence back, routes by score, gives James a usable next action, and survives a full test without manual data repair.
English