Show Notes
The single strongest predictor of a signed lease was not the special, not the square footage, not a scarcity tactic. It was whether the leasing agent laughed on the phone.
Funding stage: Seed. Grotto AI closed a $10 million seed led by ICONIQ in February 2026, with the company about one year old. Clear Seed, not Series A Plus.
There is thirty billion dollars a year in lost rent sitting in empty units across America, and that vacancy quietly erases roughly half a trillion dollars of property value. Everyone assumed the fix was price, amenities, or a slicker chatbot. Then Nick Deveau and his co-founder Ben Epstein got their hands on millions of real leasing calls from one of the largest apartment owners in the country, pointed a team of machine learning engineers at the data, and found something nobody scripted for. The second strongest predictor of a signed lease was whether the agent asked a genuinely curious question. The first was simply whether they laughed.
So Grotto AI did the counterintuitive thing. While most of the industry raced to replace humans with voice agents, Grotto built a tool to make humans better at the one thing only humans can do: build rapport. A leasing agent gets a push notification fifteen minutes before a tour telling them the prospect has a dog named Fido, loves natural light, and drives a Subaru. They record the tour on a small clip-on mic, get instant feedback on what they crushed and what they missed, and Grotto drafts the personalized follow-up, catches the special they forgot to mention, and quietly does the CRM grunt work. Nick calls it targeted advertising for the real world. Ryan called it a second brain for the field. Both are right.
This episode is the clearest case study going for vertical AI: pick one painful, measurable leak, capture data nobody else has, and sell revenue instead of cost cuts.
Named Frameworks
The Last Two Yards
Vibe-coding 80% of a product in twenty minutes is the easy, sexy part. The last two yards are where most AI dies.
- →Deployment, change management, and getting every end user bought in is where it actually breaks.
- →Classic 80/20: the final 20% is 80% of the work, and it is exactly what the big labs are now staffing with armies of forward-deployed people.
Sensor-First Data Moat
Put a sensor where one never existed.
- →A microphone in a 45-minute leasing tour captures roughly 100x the information anyone ever had on a prospect.
- →That proprietary data unlocks both human-plus-AI moments and fully agentic workflows.
- →It is genuinely defensible because nobody else has it.
Laughter Over Logic
A bottoms-up analysis of millions of calls found relationship signals beat tactics.
- →Laughter lifted conversion 48%, genuine curiosity 35%, and standard sales asks only about 14%.
- →The takeaway: coach for warmth, not for scripts.
Sell Revenue, Not Reduction
Most AI pitches sell labor or OpEx reduction, which is bounded.
- →Grotto sells more rent revenue, which is unbounded.
- →That is why the buyer leans in instead of bracing for layoffs.
Hire for Rapport, Train the Skill
Because warmth is the differentiator, you can hire anyone who naturally connects.
- →Hire bartenders, retail and hospitality people, anyone who builds rapport, then compress ramp from 90 days to 9 with the tool.
- →The talent pool widens and turnover hurts less.
Configuration to Customization Flip
In 2016 the engineering gospel was 'configure, never customize.' That has flipped.
- →As the cost of building collapsed, last-mile custom code became the thing that makes you sticky.
- →The build-versus-buy debate is shifting toward buy because nobody wants to maintain it forever.
Voice of the Customer Go-To-Market
Cold outbound dies in tight verticals.
- →This industry is conference-driven and relationship-led.
- →The motion is industry advisors at the table plus customers who become your best salespeople on stage.
Founder Experiment
Run the Laughter Audit on your own funnel. Record your next ten sales or discovery calls, then tag each one for two things only: moments of real laughter and genuinely open, curious questions. Map those tags against which calls converted. If warmth out-predicts your pitch, you just found the thing to coach. If it does not, you found out your funnel is different, which is also worth knowing.
Glossary
Vacancy loss
Rent revenue lost to empty units, roughly $30B a year, tied to about $500B in erased property value.
NOI
Net operating income, the bottom-line number Grotto ties its results to.
Forward-deployed engineer
The Palantir-style model of embedding engineers to build last-mile custom solutions for a specific customer.
Change management
The human work of getting an org to actually adopt new tech, not just buy it.
Design partner
An early customer who builds the product hand in hand with you.
LOI
Letter of intent, the early signal Nick groveled for to prove demand.
ICP
Ideal customer profile.
Uncanny valley
The creepy zone where an AI feels almost human and people recoil once they realize it is not.
Prosody
Tone, rhythm, and intonation of speech, part of the 300 to 500 features Grotto scores per call.
Vertical AI
AI purpose-built for one industry rather than horizontal and general.
RFP
Request for proposal, the competitive bid Grotto wins by being leasing-specific.
Agentic workflow
A task an AI agent completes on its own without a human in the loop.
Q&A: What Founders Ask After This Episode
What actually closes an apartment lease?
Rapport. In Grotto's analysis of millions of calls, laughter was the top predictor of conversion, ahead of price, features, or scarcity tactics.
Why is the last mile of AI deployment so hard?
Building the product is now fast, but adoption, change management, and custom last-mile integration are 80% of the real work, which is why even frontier labs staff it heavily.
Should a founder build a narrow vertical AI product or a horizontal one?
Being purpose-built for one vertical let Grotto win RFPs and build functionality no horizontal competitor could match.
How do you sell AI into a legacy industry without triggering job-loss fear?
Sell revenue, not labor reduction, and lead implementation by telling end users the tool is built to get them promoted and paid more, not replaced.
How do you build a data moat?
Put a sensor where one never existed, in Grotto's case a microphone in a leasing tour, to capture proprietary data nobody else has.
What is vacancy loss?
Roughly $30 billion a year in lost rent across the US, tied to about $500 billion in erased property value.
How do you go to market in a tight, relationship-driven industry?
Skip cold outbound. Use industry advisors and customer voice, because everybody knows everybody and reputation travels.
Five Founder Questions This Episode Answers
- →How do I get AI actually adopted by frontline teams, not just approved by the execs who signed the contract?
- →Should I build a narrow vertical product or a horizontal one, and which is more defensible right now?
- →How do I sell AI into a fearful, legacy industry without the end users assuming I am there to take their jobs?
- →What is the right go-to-market when cold outbound does not work and the whole industry runs on relationships?
- →How do I capture proprietary data that competitors simply cannot get?
URLs Mentioned in the Episode
- Grotto AIhttps://grotto.ai
- Nick Deveau on LinkedInhttps://www.linkedin.com/in/nick-deveau-a6241379/
- Ryan Estes on LinkedInhttps://www.linkedin.com/in/estesryan/
- AI for Foundershttps://aiforfounders.co
- Inbox Alchemyhttps://inboxalchemy.co
- Robin Hood Foundationhttps://robinhood.org
