Growth lessons from consumer tech operator‑investor
- B2B
- Go-to-Market
- Marketing Consultant
- Entrepreneurship, Artificial Intelligence, Market Intelligence
Most vertical AI companies fail because they're building features, not businesses. Rachel ten Brink, GP at Red Bike Capital and former Scentbird co-founder (scaled to 500K+ subscribers, $29M raised), breaks down how to build defensible vertical AI that survives the regulatory gauntlet. She reveals her operator's test for distinguishing product businesses from services wrappers, explains how winning data network effects emerge from proprietary processing of public data, and shares the specific go-to-market playbook that works for enterprise AI sales cycles.
About the speaker
Rachel ten Brink
Redbike Capital
- Part 1 Growth lessons from consumer tech operator‑investor
Episode Chapters
-
02:15: Product versus services wrapper
Real vertical AI companies solve specific problems with automation, not consulting disguised as software with 70% human customization.
-
04:59: Customer love beats revenue
Early validation isn't about fundraising or even revenue—it's finding customers who stick through product bumps because you're solving something critical.
-
06:16: Data flywheels by design
Wi ers embed proprietary data loops from day one, making their product stickier as users contribute more private information that enhances value.
-
11:09: Go narrow or go broke
Vertical AI founders must pick one core benefit and ruthlessly prioritize—trying to market multiple value propositions kills focus and burns cash.
-
16:37: Do unscalable things first
Companies are like children—don't act like an adult when you're still a toddler, but design handmade solutions that can eventually shift to automation.
-
19:13: Get anthropological about users
Stop dwelling in abstractions and jazz hands—put on safari hats, record exact customer words, and understand where they eat breakfast.
-
22:29: Scaling breaks in predictable ways
Watch for humans-in-the-loop creeping back into supposedly automated workflows—it's an early warning that margins are about to collapse.
-
26:52: Fire bad customers early
Easy-to-acquire customers are often unprofitable for a reason—sometimes you need to take a revenue step back to move forward with discipline.
-
35:53: Execution wrong, market right
Scentbird nearly died with 112 website visitors before pivoting from try-before-buy to travel sprays—knowing your market matters more than perfect execution.
-
Episode Summary
-
Why 90% of Vertical AI Startups Are Building Services Companies in Disguise
The Brutal Reality Check
Most vertical AI startups are drowning in customization quicksand while pretending they're building scalable software. Rachel ten Brink, who scaled Scentbird to 500K subscribers and now backs AI founders at Red Bike Capital, drops the uncomfortable truth: "Are you building something that can be replicable, can service multiple customers, or are you going to have to come in and it's basically 30% product and then 70% people customizing?" The answer determines whether you're building a venture-scale business or an overengineered consulting firm. -
The Post-Check Playbook
The Three-Part Wrapper Test
Ten Brink's framework cuts through the AI hype with surgical precision. First, apply the Sam Altman test: "Every time the model gets better, does your business become more valuable or less valuable?" If OpenAI's next release threatens your existence, you're building on quicksand. Second, examine your service footprint—if you need heavy customization for each customer, you're not building software. Third, assess true scalability potential. Can this genuinely reach thousands of customers, or will you be stuck in services hell forever? -
Data Moats That Actually Matter
Forget the generic "data flywheel" pitch deck slides. Ten Brink shares how one portfolio company built defensibility using public insurance data that companies were forced to publish under the No Surprises Act. The twist? Insurance companies made the data intentionally unusable. The startup synthesized this mess into actionable intelligence, then layered on proprietary arbitration workflow data. Result: a moat built from public data processed in a way that creates net new proprietary value. -
The GTM Reality for Vertical AI
Message-Market Fit Before Product-Market Fit
Ten Brink learned this lesson scaling Scentbird: "We had several theories and they all seemed beautiful... The point is we couldn't market all of them." For vertical AI, this means obsessing over your ICP's actual buying criteria—not your nine value propositions. Spend weeks, not months, nailing one core benefit that matters to the person who writes checks. Everything else is noise that dilutes your message and confuses your sales process. -
The Unscalable Things That Scale
"If you don't do zero to one, one to ten, ten to a hundred, you don't get to a thousand." Ten Brink advocates for strategic handmade solutions early on—with one critical caveat: never build something that can't eventually shift from manual to automated. One founder in her portfolio closes enterprise logos by hopping on planes for in-person meetings. Completely unscalable? Yes. Building the customer love and word-of-mouth that creates compound growth? Absolutely. -
When Scale Breaks Everything
The Two-Week Product Sprint Solution
When hallucinations spike or humans-in-the-loop multiply beyond projections, Ten Brink prescribes radical medicine: "Take a break on onboarding customers for two weeks and really address what are the product issues." This isn't about perfection—it's about preventing the death spiral where you're drowning in demand but hemorrhaging margins because every customer needs hand-holding. Set hard deadlines, communicate transparently with customers, and fix the core issues before they kill your unit economics. -
Fire Your Bad Customers
The hardest lesson: "Bad customers come in different flavors." Some are too small to afford you. Others are large enterprises using you as free R&D while their internal teams build a copycat. Ten Brink's advice is blunt: either create a completely touchless fu el for low-value segments or cut them loose entirely. Brex famously fired their SMB customers to focus upmarket. Sometimes the easy-to-acquire customers are easy precisely because they're profit destroyers. -
The Unfair Advantage
Building vertical AI that scales requires accepting three uncomfortable truths. First, if your AI product needs constant customization, you're building a services company with a tech veneer. Second, real defensibility comes from turning public or workflow data into proprietary intelligence—not from using the latest model. Third, sustainable growth means saying no to bad customers and fixing product issues even when demand is exploding. The wi ers understand that in vertical AI, discipline beats growth theater every time.
Up Next:
-
Part 1Growth lessons from consumer tech operator‑investor
Most vertical AI companies fail because they're building features, not businesses. Rachel ten Brink, GP at Red Bike Capital and former Scentbird co-founder (scaled to 500K+ subscribers, $29M raised), breaks down how to build defensible vertical AI that survives the regulatory gauntlet. She reveals her operator's test for distinguishing product businesses from services wrappers, explains how winning data network effects emerge from proprietary processing of public data, and shares the specific go-to-market playbook that works for enterprise AI sales cycles.