🎯 MISSION BRIEFING
THE PLAY
Objective: Sound genuinely interested (we ARE interested in what they've built). Ask deep technical questions.
Extract: voice model, ST integration method, booking logic, speed-to-lead architecture, widget technology.
Cover story: "We're excited about the demo Stephanie saw. I'm the tech guy, I need to understand the integration with our systems before we commit."
Ashton's role: Ask questions from a CSR perspective. "How would this work with my current workflow?" This is 100% genuine -- he IS the one who would use it.
DO NOT: Mention Nexus, Daniel, Retell, or that you're building a competitor. You're "evaluating" their product.
🧠 INTEL FROM FIRST DEMO (What We Already Know)
$3,500/mo
Enterprise Price
CONFIRMED FACTS FROM DEMO 1
✅ Enterprise plan: $3,500/mo -- inbound, outbound, speed-to-lead, scheduling widget
✅ 5,000 minutes included, $0.70/min overage
✅ 6-month contract (not 12), can downgrade mid-term, no penalty
✅ No onboarding fee, first bill 30 days after signing
✅ QA/coaching add-on: $3,698/mo total (Stephanie skipping this)
✅ "Dane" voice has highest booking rate (cloned real voice, not synthetic)
✅ Custom voice cloning: "not available yet, could be engineering request"
✅ Spanish AI agent available
✅ Brenda Barker (Barker and Sons): 62 calls, 71.43% booking rate, $15K revenue in 30 days on inbound alone
✅ Widget: 28 booked appointments, $28K revenue, 62% booking rate
✅ "Integrates directly with ServiceTitan -- not API or manual, everything syncs automatically"
✅ Outbound: charges only if recipient picks up, billed to the second
✅ Text conversations: flat 2 min talk time (~$1.40/conversation)
✅ Spam numbers auto-blocked, can pre-load known spam
✅ Job priority configurable (sewer $9K over drain $1K)
✅ Booking rate insights explain why calls don't book
✅ Onboarding backlog -- go-live target: end of first week of April
✅ Gerard's number: 305-766-5098
📋 STEPHANIE'S FEEDBACK ON DANIEL (Our Current AI)
PROBLEMS TO SOLVE (Stephanie's exact words)
🔴 Doesn't NOW auto-books via web scheduler automation (built March 31 2026). Playwright fills ST form automatically.
🔴 Sounds "a little bit too slow" -- cadence is off
🔴 Not answering after 3 rings
🔴 "Sounded really weird" -- needs to be close to Dane voice quality
🔴 Needs to prioritize better-selling jobs
🔴 Needs Ashton's cadence and voice characteristics
🔴 Check Retell/3CX contract for voice cloning restrictions
🟡 Find YouTube examples of really good customer service
🟡 Full integration -- re-engineer 10x better
🔍 INTELLIGENCE EXTRACTION QUESTIONS
🎯 TECHNICAL ARCHITECTURE (Ask naturally, don't interrogate)
👤 ASHTON'S QUESTIONS (CSR perspective -- 100% genuine)
A1: "If I'm on a call and another call comes in, does the AI pick it up automatically? Or do I have to toggle something?"
Reveals: overflow routing logic, 3CX/PBX integration method
A2: "Can I see what the AI booked in real-time? Like does it show up in ST immediately or is there a delay?"
Reveals: sync latency, real-time vs batch booking
A3: "What happens if the AI books a time slot that's already full? Does it know our availability?"
Reveals: availability awareness -- do they read ST dispatch capacity or just book blindly?
A4: "Can I train it on specific things? Like if someone asks about our Bright Club membership, will it know what to say?"
Reveals: custom knowledge base, RAG architecture, training process
A5: "How does the AI handle a customer who's upset? Like a callback on a bad job?"
Reveals: sentiment handling, escalation logic, edge case management
🔨 WHAT WE CAN ALREADY DO (Morpheus/Daniel)
ALREADY BUILT OR TRIVIAL TO BUILD
✅ Voice agent: Daniel on Retell AI -- $30-40/mo vs $3,500/mo
✅ ST customer lookup: We have the API, 5,392 customers in our DB
✅ ST booking: POST /jpm/v2/jobs -- we can create bookings via API
✅ Call recording + transcripts: Retell provides these automatically
✅ Spanish: Retell supports multi-language agents
✅ Speed-to-lead: Our sync daemon + Retell outbound = sub-60-second response
✅ Job priority: GPT-4o function calling classifies intent in real-time
✅ Widget: ST web scheduler already on 21 pages
✅ After-hours routing: 3CX IVR 821 -> Queue 810 -> Daniel
✅ Overflow: 3CX ring group falls through to Daniel on busy
✅ Multiple simultaneous calls: Retell handles unlimited concurrent
✅ Analytics dashboard: Morpheus already has call stats
GAPS TO CLOSE (What Broccoli has that we don't yet)
🟡 Voice quality: Daniel needs cadence improvement (Stephanie said "too slow"). Retell has voice cloning -- clone Ashton's voice?
🟡 Answer speed: Not picking up within 3 rings. Fix: adjust 3CX queue timeout
🟡 ST booking via API: We have the API but haven't built the booking flow in Daniel's function calls yet
🟡 Availability awareness: Need to check ST dispatch capacity before booking
🟡 Booking rate analytics: Need to track why calls don't book (their "excused calls" concept)
🟡 Outbound happy calls: Retell can do this, just needs to be built
🟡 Proven track record: Broccoli has Brenda's numbers. We have zero production calls.
💰 COST: BROCCOLI vs MORPHEUS/DANIEL
🎭 MEETING STRATEGY (1:30 PM - 2:30 PM)
MINUTE-BY-MINUTE PLAN
0-5 min: Warm intro. "Hey Gerard, thanks for doing this again. I'm Robert, the tech guy. Stephanie brought me in because I need to understand the integration side before we commit. Ashton is here too -- he's our lead CSR, this directly affects his workflow."
5-15 min: Let Gerard do his demo. Don't interrupt. Take notes on EVERYTHING he shows on screen. Screenshots if possible.
15-30 min: Technical questions (Q1-Q7 above). Frame as: "I'm the integration guy, I need to make sure this plays nice with our existing stack."
30-40 min: Ashton's questions (A1-A5). Let Ashton drive. His questions are 100% genuine AND extract valuable intel.
40-50 min: Ask about the dashboard/analytics. "Can you show us the Barker dashboard again? How does the revenue attribution work?"
50-55 min: "We're very interested. Let us huddle with Kalen tonight and get back to you by end of week." This buys time.
55-60 min: Ask for a RECORDING of the demo. "Can you send us a recording? We want to share with Kalen who couldn't be here." This is gold -- we can re-watch and extract every detail.
⚠ RED FLAGS TO WATCH FOR
🔴 "Integrates directly with ST -- not API"
This claim from Demo 1 is suspicious. ST only offers API integration. If they say "partner integration" or "embedded" -- that's different. Press on this.
🔴 Booking rate claims without context
71.43% booking rate sounds great, but what counts as a "booking"? If it's just scheduling a call, that's easy. If it's a confirmed job in ST with customer verified, that's impressive.
🔴 "Dane is a cloned real voice"
If they use ElevenLabs or PlayHT, we can achieve the same quality. Ask: "How realistic is it? Can we hear a sample call?" Then compare to Retell's options.
🔴 Contract pressure
Gerard may push to close on the call. Standard response: "We need to bring this to Kalen. He's the financial decision maker." This is true AND buys time.
📝 AFTER THE CALL: ACTION ITEMS
IMMEDIATE (Within 24 hours)
1. Document every technical detail revealed in the call
2. Update BSP_Proprietary_Stack.html with Broccoli feature map
3. Fix Daniel's cadence and speed issues on Retell
4. Adjust 3CX queue timeout for <3 ring pickup
5. Build ST booking function call into Daniel's agent
6. Clone Ashton's voice on Retell (if contractually allowed)
7. Test Daniel with improved voice vs Broccoli's Dane
8. Build "Broccoli Killer" feature parity checklist
DECISION FRAMEWORK
Sign with Broccoli IF:
• Their ST integration is truly deep (not just API wrappers we can replicate)
• Voice quality is genuinely un-replicable
• Go-live in April means revenue BEFORE we could build Daniel v2
• Use the 6 months to study their product from the inside
Build our own IF:
• Their "integration" is just ST API calls (which we already have)
• Voice quality is ElevenLabs/PlayHT (which we can access for $20/mo)
• Daniel v2 can match their booking rate within 30 days
• $41,400/yr savings funds 1000+ hours of development
Hybrid approach:
• Sign Broccoli for 6 months to get immediate coverage + study their product
• Build Daniel v2 in parallel using extracted intelligence
• Cancel Broccoli at month 6 when Daniel v2 is production-ready
• Net cost: $21,000 for 6 months of coverage + intelligence = worth it if revenue recovery is $25K+/mo
📱 QUICK REFERENCE (Keep this open during the call)
Gerard Brown: Sales rep, 305-766-5098
Brenda Barker: Barker and Sons (Stephanie's mother-in-law), already on Broccoli
Price: $3,500/mo enterprise, 5,000 min, $0.70 overage, 6-mo contract
Onboarding: 3 days, go-live end of first week of April
Products: Inbound + Outbound + Speed-to-Lead + Widget (QA is add-on)
Voice: Dane (highest booking rate, cloned real voice, 50+ options)
BSP CSR hours: 8 AM - 5 PM, peak 6-10 AM (2 hrs uncovered)
BSP avg ticket: $2,600, sewers $9K-$40K
Our AI: Daniel on Retell, $30-40/mo, needs voice/speed fixes
Key phrase: "We need to run this by Kalen" (buys time)
BLIND: We haven't heard their AI handle a real call
Gerard showed dashboards and numbers but did we actually HEAR Dane answer a live call? Ask for a live demo call during this meeting. "Can we call the Barker number right now and hear it?"
BLIND: We don't know their failure rate
71.43% booking rate means 28.57% of calls DON'T book. What happens to those? Do they get a callback? Are they lost? What's the "excused" vs "failed" breakdown? This is critical.
BLIND: No data on caller satisfaction
Do callers know they're talking to AI? Do they care? Has any customer complained? Stephanie said Daniel "sounded really weird." Broccoli's Dane might sound great on demo but how do real customers react?
BLIND: We don't know their uptime/reliability
What happens when Broccoli's servers go down? Does it fail to voicemail? Is there an SLA? Our 3CX Jessica crashed after 1 call. What's their track record?
WEAK: We haven't tested Daniel against a real customer
Daniel works technically but zero production calls. We can't claim our solution is better until we have data. Broccoli has Brenda's numbers. We have zero.
WEAK: Stephanie already verbally committed
From transcript: "Stephanie verbally committed to signing." This creates social pressure. Gerard will reference this. Need graceful way to pause without burning the bridge.
WEAK: Brenda is Stephanie's mother-in-law
This is a family referral. Rejecting Broccoli could create family tension. The "we need more time to evaluate" approach is safer than "we're building our own."
WEAK: 6 AM-8 AM gap is REAL revenue loss
Peak call window is 6-10 AM but CSRs start at 8 AM. That's 2 hours of peak time with zero coverage every day. This IS the problem Broccoli solves and we need to solve it too.
✅ Brenda's $15K claim is verifiable. Ask Gerard: "Can we see the actual ST revenue report tied to Broccoli-booked jobs?" If yes, the number is real.
✅ $3,500/mo pricing confirmed. Matches the transcript verbatim. Enterprise plan, includes all 4 products.
❌ "Not API or manual" integration claim unverified. ST only offers API access. This claim needs deep technical probing. Could be marketing language for "our middleware uses the API."
❌ 71.43% booking rate context unknown. Is this ALL calls or just "bookable" calls? What counts as a booking? How many were "excused"? The denominator matters more than the percentage.
⚠ $28K widget revenue attribution unclear. Did the widget CREATE these leads or were they existing customers who would have called anyway? Attribution matters for ROI calculation.
⚠ BSP's after-hours call volume is estimated, not measured. We said "3-5 after-hours calls/day" but we don't actually track this. 3CX CDR was just enabled. Get real numbers before committing $42K/yr.
✅ Our cost comparison is accurate. Retell: $0.10-0.15/min. 5,000 min = $500-750/mo max vs $3,500/mo. Savings: $2,750-3,000/mo verified.
✅ Retell voice quality is competitive. Retell uses ElevenLabs-tier voices with sub-1-second latency. Voice cloning available. Technically capable of matching Dane.
STRENGTHS (Signing Broccoli)
• Immediate coverage (go-live early April)
• Proven booking rate with Brenda's data
• ST integration already built and tested
• Spanish support day one
• 50+ voice options including Dane
• Professional support + biweekly CSM meetings
• Brenda referral = family trust
WEAKNESSES (Signing Broccoli)
• $3,500/mo = $42K/yr when our build costs $600/yr
• 6-month lock-in minimum
• Another vendor dependency (already fighting ST dependency)
• Their data sits in THEIR system, not Morpheus
• No custom voice cloning (Ashton's voice)
• Can't modify behavior -- stuck with their logic
• Overage at $0.70/min adds up fast at scale
OPPORTUNITIES (Building Our Own)
• $41,400/yr savings (funds 1000+ dev hours)
• Full integration with Morpheus + Trust Engine
• Custom voice cloning (Ashton's actual voice)
• Revenue attribution via Trust Engine Eq1
• 100 Year Plumbing on same system ($1/mo extra)
• Sell to other plumbers (productize like Broccoli did)
• Speed-to-lead feeds our outreach engine
• Study Broccoli from inside if we sign short-term
THREATS (Building Our Own)
• Time to production: Daniel v2 is 2-4 weeks away
• Every day without coverage = lost after-hours leads
• Stephanie already verbally committed to Broccoli
• Family dynamics (Brenda referral)
• Daniel v1 feedback was negative ("sounded weird")
• Building while also building ST Killer = resource split
• If Daniel fails in production, BSP loses trust in Morpheus
• Broccoli has 100+ clients. We have 0 production calls.
Phase 1: Voice Selection (Week 1)
• Test 3 Retell voices: Default male, cloned Ashton, "Dane-like" warm male
• Internal team calls 10 each. Rate on: naturalness, trustworthiness, speed
• Measure: which voice do Stephanie/Ashton/Kalen prefer?
• Winner becomes Daniel v2's default voice
Phase 2: Greeting + Cadence (Week 2)
• Test A: "Thank you for calling Bright Side Plumbing, this is Daniel, how can I help you today?"
• Test B: "Bright Side Plumbing, what's going on today?" (Ashton's style -- direct, warm)
• Test C: "Hey there, you've reached Bright Side Plumbing. What can we do for you?"
• Measure: caller engagement, time-to-booking, hang-up rate
Phase 3: Soft Launch (Week 3)
• After-hours only (9 PM - 8 AM). Zero risk to daytime operations
• Track: calls answered, booking rate, call duration, customer callback rate
• Compare to Broccoli's claimed 71.43% booking rate
• Target: 50%+ booking rate in week 1, 65%+ by week 4
Phase 4: Full Deploy (Week 4+)
• Add overflow during business hours (when Ashton/Jordan are on calls)
• Add 6 AM-8 AM pre-shift coverage
• Add speed-to-lead (respond to web forms in <60 seconds)
• Measure full funnel: call > book > dispatch > complete > invoice > revenue
Q8: "What's your typical customer churn rate? How many companies cancel after 6 months?"
WHY: If churn is high, the product may not deliver long-term value. If low, it works.
Q9: "Has any customer built their own solution and left? What did they struggle with?"
WHY: This reveals their moat. If nobody leaves to build their own, either the product is genuinely hard to replicate OR customers don't have the tech capability. We do.
Q10: "Can we get a trial with real calls before signing? Even 1 week at reduced rate?"
WHY: Real data > demos. If they say no, they're hiding something. If yes, we get free intelligence.
Q11: "What happens to our data if we cancel? Can we export call recordings and transcripts?"
WHY: Data portability is critical. If they lock our data, that's a red flag and a vendor dependency trap.
Q12: "You said the AI prioritizes higher-value jobs. Can you show us how that's configured? Can we see the priority settings screen?"
WHY: This reveals their intent classification UX. Screenshot the settings screen if possible.
Q13: "For the speed-to-lead, what's the actual average response time? Not the target, the measured average across all your clients."
WHY: "Under 60 seconds" is a marketing claim. Actual measured P95 latency is the real number.
Q14: "Does the AI ever mention it's AI? Like if a customer asks 'are you a real person,' what does it say?"
WHY: Some states require AI disclosure. Also reveals their prompt engineering approach.
Q15: "Can we set up a test call right now? I want to hear it handle a sewer backup emergency at 2 AM."
WHY: The most valuable 60 seconds of the entire meeting. Record this. Transcribe it. Reverse-engineer the entire conversation flow.