# AI Sales Brain - Master Plan **Status:** Planning **Created:** Jan 17, 2026 **Goal:** Unify Content Hub + Call System + Learning into one intelligent system that knows what works and gets smarter every day --- ## The Vision **Current State:** Tools sitting next to each other **Target State:** One AI brain that: - Knows what works for each lead type - Recommends the right script/message at the right time - Learns from every interaction - Surfaces patterns to the user - Gets smarter every day --- ## Architecture ``` ┌─────────────────────────────────────────────────────────────────┐ │ AI SALES BRAIN │ ├─────────────────────────────────────────────────────────────────┤ │ │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ │ │ CONTENT │ │ CALL │ │ LEARNING │ │ │ │ HUB │◄─┼─► SYSTEM │◄─┼─► ENGINE │ │ │ └──────────────┘ └──────────────┘ └──────────────┘ │ │ │ │ │ │ │ └────────────────┼─────────────────┘ │ │ │ │ │ ▼ │ │ ┌──────────────────────┐ │ │ │ INTELLIGENCE API │ │ │ │ │ │ │ │ • What works now? │ │ │ │ • Best for this lead│ │ │ │ • Market intel │ │ │ │ • Pattern detection │ │ │ └──────────────────────┘ │ │ │ │ │ ┌────────────────┼────────────────┐ │ │ │ │ │ │ │ ▼ ▼ ▼ │ │ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ │ │ PERPLEXITY │ │ GPT-4o │ │ PATTERNS │ │ │ │ Market Intel│ │ Reasoning │ │ Database │ │ │ └─────────────┘ └─────────────┘ └─────────────┘ │ │ │ └─────────────────────────────────────────────────────────────────┘ ``` --- ## The Intelligence API One endpoint that powers everything: `POST /api/ai/intelligence` ```javascript // Request { "context": "pre_call" | "during_call" | "sms_compose" | "dashboard", "lead_id": "uuid", "request": "recommend_script" | "what_to_say" | "market_intel" | "whats_working" } // Response { "recommendation": { "script_id": "uuid", "script_name": "Reactivation Check-in", "confidence": 0.87, "reason": "43% success with similar leads (plumber, 3 weeks quiet)" }, "market_intel": { "demand_trend": "+23%", "jobs_in_area": 8, "competitor_news": "HiPages raised fees" }, "patterns": { "best_time": "Tuesday 10am", "best_approach": "Mention jobs in area", "avoid": "Don't lead with pricing" }, "similar_wins": [ { "lead": "Mike's Plumbing", "approach": "Reactivation script + jobs angle", "days_to_convert": 3 } ] } ``` --- ## Data Model: What We Track ### 1. Script/Template Performance ```sql CREATE TABLE content_performance ( id UUID PRIMARY KEY, content_type VARCHAR(20), -- 'script', 'sms_template', 'objection_response' content_id UUID, -- Usage times_used INTEGER DEFAULT 0, -- Outcomes success_count INTEGER DEFAULT 0, callback_count INTEGER DEFAULT 0, conversion_count INTEGER DEFAULT 0, -- Segmented performance performance_by_trade JSONB, -- {"plumber": {"used": 50, "success": 23}, ...} performance_by_day JSONB, -- {"monday": {"used": 30, "success": 15}, ...} performance_by_hour JSONB, -- {"10": {"used": 20, "success": 12}, ...} performance_by_lead_temp JSONB, -- {"hot": {...}, "warm": {...}, ...} -- Trends last_7_days_success_rate DECIMAL, last_30_days_success_rate DECIMAL, trend VARCHAR(10), -- 'up', 'down', 'stable' updated_at TIMESTAMP ); ``` ### 2. User Performance Patterns ```sql CREATE TABLE user_patterns ( id UUID PRIMARY KEY, user_id UUID, -- Time patterns best_call_day VARCHAR(10), best_call_hour INTEGER, best_sms_day VARCHAR(10), best_sms_hour INTEGER, -- Approach patterns best_opening_style VARCHAR(50), -- 'direct', 'rapport', 'value_lead' best_trade_vertical VARCHAR(50), avg_calls_to_convert DECIMAL, avg_touchpoints_to_convert DECIMAL, -- Strengths/weaknesses objection_handle_rate JSONB, -- {"too_expensive": 0.78, "too_busy": 0.45} updated_at TIMESTAMP ); ``` ### 3. Market Intelligence Cache ```sql CREATE TABLE market_intel ( id UUID PRIMARY KEY, region VARCHAR(50), -- 'sydney', 'melbourne', etc. trade VARCHAR(50), -- 'plumber', 'electrician', etc. -- Demand signals demand_index INTEGER, -- 0-100 demand_trend VARCHAR(10), -- 'up', 'down', 'stable' demand_change_percent DECIMAL, -- Job data active_jobs_count INTEGER, avg_job_value DECIMAL, -- Competition competitor_activity JSONB, -- Perplexity research industry_news TEXT[], hiring_trends TEXT[], searched_at TIMESTAMP, expires_at TIMESTAMP ); ``` ### 4. Pattern Library ```sql CREATE TABLE discovered_patterns ( id UUID PRIMARY KEY, pattern_type VARCHAR(50), -- 'conversion', 'timing', 'approach', 'objection' -- The pattern description TEXT, -- "Plumbers who go quiet for 3 weeks convert 43% with reactivation script" -- Conditions conditions JSONB, -- {"trade": "plumber", "days_quiet": [14, 28], "script": "reactivation"} -- Evidence sample_size INTEGER, confidence DECIMAL, success_rate DECIMAL, -- Status is_active BOOLEAN DEFAULT true, discovered_at TIMESTAMP, last_validated_at TIMESTAMP ); ``` --- ## Phase 1: Foundation (Week 1) ### 1.1 Create Intelligence API **File:** `src/routes/intelligence.js` ```javascript // Core endpoint that powers everything router.post('/intelligence', async (req, res) => { const { context, lead_id, request } = req.body; // Gather all context in parallel const [ leadData, contentPerformance, userPatterns, marketIntel, similarWins, discoveredPatterns ] = await Promise.all([ getLeadWithHistory(lead_id), getContentPerformance(context), getUserPatterns(req.user.id), getMarketIntel(leadData?.trade, leadData?.region), getSimilarWins(leadData), getRelevantPatterns(leadData) ]); // AI reasoning layer const recommendation = await generateRecommendation({ context, request, leadData, contentPerformance, userPatterns, marketIntel, similarWins, discoveredPatterns }); return res.json(recommendation); }); ``` ### 1.2 Create Performance Tracking **File:** `src/services/performanceTracker.js` Track every interaction: - Script used → outcome - SMS sent → reply/conversion - Objection response → worked/didn't - Call time → outcome ### 1.3 Create Pattern Discovery Job **File:** `src/jobs/patternDiscovery.js` Daily job that: - Analyzes last 30 days of data - Finds statistically significant patterns - Stores in `discovered_patterns` table - Alerts when new patterns found --- ## Phase 2: Content Hub Integration (Week 2) ### 2.1 "What's Working" Dashboard Top of Content Hub shows: ``` ┌─────────────────────────────────────────────────────────────┐ │ 🧠 AI INSIGHTS - Updated 10 min ago │ │ │ │ 🔥 HOT RIGHT NOW │ │ "Reactivation Check-in" script is crushing it - 47% success │ │ this week (up from 31% last week) │ │ │ │ 📈 PATTERN DETECTED │ │ Your Tuesday 10am calls convert 2.3x better than average. │ │ You have 5 callbacks scheduled for Friday - move them? │ │ [Yes, Reschedule] [No Thanks] │ │ │ │ 💡 TRY THIS │ │ Plumbers are responding well to "jobs in area" angle. │ │ You have 3 plumber leads today - use this approach? │ │ │ └─────────────────────────────────────────────────────────────┘ ``` ### 2.2 Smart Script Recommendations When viewing scripts: ``` ┌─────────────────────────────────────────────────────────────┐ │ 📞 Reactivation Check-in │ │ │ │ ⭐ AI RECOMMENDED for your next 3 calls: │ │ • Mike's Plumbing (plumber, 3 weeks quiet) - 87% match │ │ • Sparky Solutions (electrician, callback) - 72% match │ │ • BuildRight Co (builder, re-engage) - 68% match │ │ │ │ 📊 PERFORMANCE │ │ Overall: 34% success │ │ For plumbers: 47% success ← your best trade │ │ On Tuesdays: 52% success ← your best day │ │ At 10am: 61% success ← sweet spot! │ │ │ │ 🏆 SIMILAR WINS │ │ "Used this with Mike's Plumbing last week - converted in │ │ 2 calls after mentioning the 8 jobs in his area" │ │ │ └─────────────────────────────────────────────────────────────┘ ``` ### 2.3 AI Compose with Context When composing SMS: ``` ┌─────────────────────────────────────────────────────────────┐ │ ✨ AI COMPOSE │ │ │ │ Writing for: Mike's Plumbing (plumber, went quiet) │ │ │ │ 🧠 AI KNOWS: │ │ • Plumbers respond to "jobs in area" (67% reply rate) │ │ • Mike opened last SMS but didn't reply │ │ • 8 plumbing jobs posted in his area this week │ │ • Best time to send: Tuesday 10am (your pattern) │ │ │ │ [Generate Smart Message] │ │ │ │ RESULT: │ │ "Hey Mike, saw 8 plumbing jobs come through in your area │ │ this week - keen to send them your way? No stress if │ │ you're sorted 👍" │ │ │ │ Predicted reply rate: 34% (above your average of 18%) │ │ │ └─────────────────────────────────────────────────────────────┘ ``` --- ## Phase 3: Call System Integration (Week 3) ### 3.1 Smart Intel Brief Replace current brief with AI-powered version: ``` ┌─────────────────────────────────────────────────────────────┐ │ 🧠 INTEL BRIEF - Mike's Plumbing │ │ │ │ ═══ AI RECOMMENDATION ═══ │ │ │ │ USE: "Reactivation Check-in" script │ │ CONFIDENCE: 87% │ │ │ │ WHY: Similar leads (plumber, 3 weeks quiet) convert at 43% │ │ with this script. Mike downloaded pricing guide but never │ │ called - mention that. │ │ │ │ ═══ MARKET INTEL (Live) ═══ │ │ │ │ 📈 Plumbing demand UP 23% in Sydney this week │ │ 💼 8 plumbing jobs posted in Mike's area │ │ 📰 News: Plumber shortage hitting Sydney's west │ │ │ │ USE THIS: "Heaps of work coming through - just this week │ │ we've got 8 plumbing jobs in your area" │ │ │ │ ═══ YOUR PATTERNS ═══ │ │ │ │ ✅ Tuesday 10am = your best time (52% success) │ │ ✅ "Jobs in area" angle works for you (67% reply) │ │ ⚠️ Avoid leading with pricing (drops success 30%) │ │ │ │ ═══ SIMILAR WIN ═══ │ │ │ │ 🏆 Last week: "Converted Pete's Plumbing with same │ │ approach - mentioned jobs in area, he signed up same │ │ day. 3 calls total." │ │ │ │ ═══ PREDICTED OBJECTIONS ═══ │ │ │ │ 🛡️ "Too busy" (72% likely) │ │ → Your response success rate: 78% │ │ → "Totally get it - that's actually why tradies love │ │ us. Post when you need workers, we handle the rest." │ │ │ │ [Ready to Call] │ └─────────────────────────────────────────────────────────────┘ ``` ### 3.2 Live Call Intelligence During call, AI assistant shows: ``` ┌─────────────────────────────────────────────────────────────┐ │ 🎯 LIVE ASSIST │ │ │ │ CURRENT STAGE: Discovery │ │ ████████░░░░░░░░ 45% │ │ │ │ ┌─────────────────────────────────────────────────────────┐ │ │ │ 💡 AI DETECTED: He mentioned "too many calls" │ │ │ │ │ │ │ │ THIS WORKS (73% success): │ │ │ │ "I hear you - that's actually why we built RateRight │ │ │ │ different. You post when YOU need workers, no cold │ │ │ │ calls from us." │ │ │ │ │ │ │ │ [Copy] [Used It 👍] [Didn't Work 👎] │ │ │ └─────────────────────────────────────────────────────────┘ │ │ │ │ ┌─────────────────────────────────────────────────────────┐ │ │ │ 🔥 BUYING SIGNAL: "How many workers do you have?" │ │ │ │ │ │ │ │ HE'S INTERESTED! Close with: │ │ │ │ "Great question - we've got 200+ verified tradies in │ │ │ │ your area. Want me to show you who's available for │ │ │ │ your next job?" │ │ │ │ │ │ │ │ [Copy] [Closing Now...] │ │ │ └─────────────────────────────────────────────────────────┘ │ │ │ │ 📊 LIVE STATS │ │ Call duration: 4:32 │ │ Sentiment: Positive (warming up) │ │ Conversion probability: 67% ↑ │ │ │ └─────────────────────────────────────────────────────────────┘ ``` ### 3.3 Post-Call Learning After every call: ``` ┌─────────────────────────────────────────────────────────────┐ │ 📊 CALL COMPLETE - Learning from this call │ │ │ │ Duration: 8:34 │ │ Outcome: Callback scheduled │ │ │ │ 🧠 AI LEARNED: │ │ │ │ ✅ "Jobs in area" angle → led to callback │ │ (Adding to your success patterns) │ │ │ │ ✅ "Reactivation Check-in" script → worked │ │ (Success rate now 48% for plumbers) │ │ │ │ ✅ Objection "too busy" → handled successfully │ │ (Your handle rate now 79%) │ │ │ │ 📈 Pattern Strengthened: │ │ "Plumbers who go quiet 2-4 weeks respond well to │ │ jobs-in-area angle" (now 47 data points) │ │ │ │ [Done] │ └─────────────────────────────────────────────────────────────┘ ``` --- ## Phase 4: Market Intelligence (Week 4) ### 4.1 Scheduled Perplexity Jobs Daily job that researches: - Trade demand by region - Industry news - Competitor activity - Hiring trends **File:** `src/jobs/marketIntelligence.js` ```javascript // Run daily at 6am async function refreshMarketIntel() { const trades = ['plumber', 'electrician', 'carpenter', 'builder', 'concreter']; const regions = ['sydney', 'melbourne', 'brisbane', 'perth']; for (const trade of trades) { for (const region of regions) { const intel = await perplexity.research(` ${trade} demand ${region} Australia 2026 - Job posting trends - Hiring activity - Industry news - Average rates `); await storeMarketIntel(trade, region, intel); } } } ``` ### 4.2 Real-Time Intel Injection During calls, surface relevant market intel: ```javascript // When lead is a plumber in Sydney const marketIntel = await getMarketIntel('plumber', 'sydney'); // Inject into call assist "📈 Market Intel: Plumbing demand up 23% in Sydney. 8 jobs posted in his area this week. Use: 'Heaps of work coming through right now'" ``` --- ## Phase 5: Pattern Discovery (Week 5) ### 5.1 Automated Pattern Detection Daily job that finds patterns: ```javascript async function discoverPatterns() { // Find timing patterns const timingPatterns = await analyzeTimingPatterns(); // "Tuesday 10am has 2.3x conversion rate" // Find approach patterns const approachPatterns = await analyzeApproachPatterns(); // "Jobs in area angle has 67% reply rate for plumbers" // Find trade patterns const tradePatterns = await analyzeTradePatterns(); // "Plumbers who go quiet 2-4 weeks convert at 43%" // Find objection patterns const objectionPatterns = await analyzeObjectionPatterns(); // "Too busy objection handled 78% successfully with X response" // Store significant patterns for (const pattern of [...timingPatterns, ...approachPatterns, ...tradePatterns, ...objectionPatterns]) { if (pattern.confidence > 0.7 && pattern.sampleSize > 20) { await storePattern(pattern); await notifyUser(pattern); // Slack/in-app notification } } } ``` ### 5.2 Pattern Surfacing Show patterns to users: ``` ┌─────────────────────────────────────────────────────────────┐ │ 🧠 AI DISCOVERED - New Pattern │ │ │ │ "Electricians who downloaded the pricing guide but didn't │ │ sign up convert at 52% when you mention 'no lock-in │ │ contracts' within the first 2 minutes." │ │ │ │ Based on: 34 similar leads │ │ Confidence: 87% │ │ │ │ You have 3 electricians matching this pattern in your │ │ call list today. │ │ │ │ [Apply to These Leads] [Show Me the Data] [Dismiss] │ └─────────────────────────────────────────────────────────────┘ ``` --- ## Phase 6: A/B Testing Engine (Week 6) ### 6.1 Automatic Variation Testing AI creates and tests variations: ```javascript // AI notices a high-performing script // Generates 2 variations to test const variations = await generateVariations(topScript, { keepCore: true, varyElements: ['opener', 'value_prop', 'close'] }); // Randomly assign to 20% of matching calls // Track performance over 50 uses each // Promote winner, retire losers ``` ### 6.2 Testing Dashboard ``` ┌─────────────────────────────────────────────────────────────┐ │ 🧪 A/B TESTS RUNNING │ │ │ │ TEST: Reactivation Script - Opener Variation │ │ Status: Running (34/50 calls) │ │ │ │ ┌─────────────────────────────────────────────────────────┐ │ │ │ A (Control): "Hey {name}, just checking in..." │ │ │ │ Success rate: 43% (18/42 calls) │ │ │ │ ████████████████████░░░░░ │ │ │ ├─────────────────────────────────────────────────────────┤ │ │ │ B (Variant): "Hey {name}, quick question..." │ │ │ │ Success rate: 51% (16/34 calls) ← LEADING │ │ │ │ █████████████████████████░░░░ │ │ │ └─────────────────────────────────────────────────────────┘ │ │ │ │ Estimated completion: 3 days │ │ Statistical significance: 72% (need 95%) │ │ │ │ [Pause Test] [End Early] [View Details] │ └─────────────────────────────────────────────────────────────┘ ``` --- ## Implementation Order ### Week 1: Foundation - [ ] Create `content_performance` table - [ ] Create `user_patterns` table - [ ] Create `market_intel` table - [ ] Create `discovered_patterns` table - [ ] Build Intelligence API endpoint - [ ] Build performance tracking service ### Week 2: Content Hub Integration - [ ] Build ContentHub.jsx with tabs - [ ] Add "What's Working" dashboard section - [ ] Add AI recommendations to script cards - [ ] Integrate smart compose for SMS ### Week 3: Call System Integration - [ ] Upgrade IntelBriefModal with AI recommendations - [ ] Add confidence scores and reasoning - [ ] Integrate market intel into pre-call brief - [ ] Add pattern display to call assist - [ ] Build post-call learning feedback ### Week 4: Market Intelligence - [ ] Build Perplexity scheduled jobs - [ ] Create market intel caching system - [ ] Integrate live market intel into calls - [ ] Build trade-specific talk tracks ### Week 5: Pattern Discovery - [ ] Build pattern discovery job - [ ] Create pattern notification system - [ ] Add pattern surfacing to UI - [ ] Build pattern management (activate/retire) ### Week 6: A/B Testing - [ ] Build variation generator - [ ] Create test assignment system - [ ] Build testing dashboard - [ ] Create auto-promotion logic ### Week 7: Personal Intel (from personal-intel-plan.md) - [ ] Extend `searchPerson()` to include personal profile (family, hobbies, interests) - [ ] Add `searchPersonalNews()` for hobby/interest-related news - [ ] Extend `lead_intel` table with personal columns - [ ] Build PersonalProfileCard component - [ ] Add conversation starters to Intel Brief - [ ] Add rapport tracking to learning engine **Personal Intel Vision:** ``` 👨‍👩‍👧‍👦 Married, 2 kids 🎣 Fishing ⛳ Golf 🏉 Rugby 💬 OPEN WITH: "Hey John - how'd the fishing trip go last weekend? Saw you posted about Lake Eucumbene." 📰 NEWS HE'D CARE ABOUT: • NRL finals start next week • Lake Eucumbene stocking report just released ``` --- ## Success Metrics | Metric | Current | Target | |--------|---------|--------| | Script recommendation accuracy | N/A | 70%+ | | Call conversion rate | ~25% | 40%+ | | SMS reply rate | ~12% | 25%+ | | Time to find right script | Manual | < 5 sec | | Pattern discovery | Manual | Automated daily | | A/B tests running | 0 | 3+ always | --- ## Files to Create ### Backend - `src/routes/intelligence.js` - Main AI endpoint - `src/services/intelligenceEngine.js` - Core logic - `src/services/performanceTracker.js` - Track everything - `src/services/patternDiscovery.js` - Find patterns - `src/services/abTesting.js` - Variation testing - `src/jobs/marketIntelligence.js` - Daily market research - `src/jobs/patternDiscovery.js` - Daily pattern analysis ### Frontend - `admin/src/pages/ContentHub.jsx` - Unified content page - `admin/src/components/content/CallsTab.jsx` - `admin/src/components/content/SMSTab.jsx` - `admin/src/components/content/ObjectionsTab.jsx` - `admin/src/components/content/SettingsTab.jsx` - `admin/src/components/ai/WhatsWorking.jsx` - Dashboard widget - `admin/src/components/ai/AIRecommendation.jsx` - Recommendation card - `admin/src/components/ai/PatternAlert.jsx` - Pattern notification - `admin/src/components/ai/ABTestDashboard.jsx` - Testing UI ### Database - `supabase/migrations/ai_sales_brain.sql` --- ## The 0.1% Difference **Before (Current):** > "Here's an intel brief. Here are some scripts. Good luck." **After (AI Sales Brain):** > "Call Mike now with the Reactivation script. Similar plumbers convert at 47% with this approach. Mention the 8 jobs in his area - that angle works 67% of the time for you. It's Tuesday 10am - your best window. Last week you converted Pete's Plumbing the same way." --- **Ready to build?**