Mock vs. actual comparison. Research identified the wrong person. Fundraise advice was solid. Knowledge-as-a-service pitch landed. 7 systemic fixes identified.
Research Intel
Time Mgmt
Fundraise Talk
Room Reading
Holdback
Research matched the wrong person
"Turn your IP into an API knowledge base. Get paid for what you know without running the agency." The only pitch angle that got genuine engagement.
One word that communicates the entire value prop. Non-technical, visual, memorable. Should be a first-line talking point for all non-technical audiences.
SAFE structure, 10% ESOP, Act 60 + transfer pricing, friends & family first. His electrolyte precedent ($2M at $5M pre via SAFE) is a concrete reference.
Never leaked Commerce Machine, Renatus/SPAC, pipeline names, burn rate, or fundraising numbers. 5/5 holdback items protected despite wrong research.
Wrong Research Profile
Entire prep built for a $3.7B wealth management CEO. Actual person is an e-commerce agency owner. Root cause: common name, no verification with introducer.
25-Minute Monologue
Three speakers, no conductor. Pitch ran 10 min over plan. Garrett didn't ask a question for 15 minutes. If they haven't asked in 5 min, you've lost them.
Pushed Rejected Angle 3x
Garrett said "not really" to agency deployment. Team asked twice more. He said "intentionally downsizing" at minute 10 — team didn't adapt until minute 65.
Jargon Overload
Truncation, coherence, prompt injection, daemons, ontology, orchestrator. Garrett uses "ChatGPT and Claude casually." Every jargon word was friction.
No Demo Shown
Had a working demo, 5-act script, and pre-flight checklist. None used. For non-technical prospects, 3 minutes of showing beats 25 minutes of describing.
5 of 12 predictions correct
1 partial match · 6 wrong
SAFE recommendation
CORRECTWould suggest closing a customer first
CORRECTAct 60 discussion would happen
CORRECTHe'd share background freely
CORRECTFollow-up call secured
CORRECTWould compare to other AI companies — said "not an expert in AI" instead
PARTIALPerson's identity — entirely wrong profile
WRONGWould ask for deliverables — said "let me think" instead
WRONGWould offer network intros — no offer made
WRONGWould be evaluating investability — was in wind-down / advice mode
WRONGWould see the platform play — didn't engage deeply enough
WRONGPersonal investment interest — not expressed
WRONGBuild Knowledge-as-a-Service prototype using Garrett's e-commerce expertise
ALEXANDER + ELY
Prep 5-minute live demo (Telegram + knowledge graph + background research)
ALEXANDER
Rebuild cheatsheet with corrected profile for follow-up call
DELPHI
Carl call tomorrow at 11 AM — apply all 7 playbook fixes
MAR 28
Lomba meeting — verify research with introducer BEFORE generating prep
WEEK 03-31
Confirm research profile with introducer before generating any prep. One Telegram message would have caught the wrong-person error.
One lead speaker per phase. Hard gate at 10 min: "What questions does this raise?" Ask about their schedule in first 2 minutes.
After Phase 1, classify prospect mode (growth / wind-down / pain / curiosity / skeptic) and select the matching pitch variant.
If prospect isn't technical, show a 3-minute demo within the first 10 minutes. Showing beats describing every time.
Auto-generate corrected profile, flag research mismatches, store MIG intelligence, update cheatsheet, create follow-up tasks.
Assign roles pre-call: who opens, who demos, who handles technical Qs, who closes. Others listen and take notes.
Three-strike rule: if the prospect declines the same angle twice, it's dead. Do not attempt a third time. Pivot to a completely different value proposition.
The prep system is mechanically excellent — polished cheatsheet, mock audio, investor deliverables, holdback strategy. The failure was upstream: bad intelligence in, beautiful nonsense out. The fundraise predictions succeeded because they were based on general pre-seed patterns, not the wrong person profile. The system's domain knowledge is sound even when person-specific intelligence fails.