YC Root AccessThis AI Startup Is Taking Over Phone Sales
CHAPTERS
- 0:05 – 0:54
Simple AI: an AI voice agent that completes sales calls end-to-end
Cat explains Simple AI’s core product: a natural-sounding voice agent that answers inbound calls and can complete the entire sales flow without human involvement. The agent can explain products, answer questions, collect shipping and billing details, and place orders.
- •“AI voice that sells” positioned as a front line for inbound phone sales
- •Handles product explanation, Q&A, and customer data collection
- •Completes the full checkout flow end-to-end with no human intervention
- •Sits in front of a company’s existing phone system
- 0:54 – 1:44
Iconic customer proof: Omaha Steaks runs its main phone line on Simple AI
Jared probes for real customer examples, and Cat highlights Omaha Steaks—an established, large U.S. brand. If you call the number on their website, Simple AI answers and can take orders.
- •Simple AI already sells across diverse categories (steak, self-storage, insurance)
- •Omaha Steaks cited as a flagship customer with 100-year history
- •Simple AI is deployed on the primary inbound number (not a pilot side-line)
- •Validates the thesis that phone sales still drive major revenue for some brands
- 1:44 – 3:39
Founders’ origin story: meeting and building software inside Y Combinator
Cat and Zach recount meeting while working at YC on internal products. They describe YC’s software infrastructure (e.g., Bookface) and why software is a “secret weapon” enabling YC to operate at scale.
- •They worked on YC’s software team (Batch Team, Bookface, other founder tools)
- •YC relies on software for applications, interviews, accounting, legal, and portfolio tracking
- •Proximity to founder stories and early tech shaped their startup ambitions
- •They saw novel AI developments early through the YC ecosystem
- 3:39 – 4:47
Catching the startup bug: deciding to build after years around founders
The conversation shifts from background to motivation: working at YC made it hard not to start a company. They cite exposure to alumni stories and early breakthroughs (including OpenAI’s YC Research roots) as a catalyst.
- •Founder inspiration from watching companies go from idea to public scale
- •Unique access to early AI progress increased conviction about a big opportunity
- •They recognized a “time is now” moment for building something significant
- •Jared frames YC’s history of successful spin-outs
- 4:47 – 5:49
Idea #1: building a “better Siri” (and why it confused users)
Cat explains their initial consumer push: a general assistant that could do many actions like ordering food or calling an Uber. They learned that wide capability sets made the product hard to understand and set expectations for.
- •Consumer-first instinct based on Cat’s prior consumer/product background
- •Attempted a multi-skill assistant that could take many real-world actions
- •Users struggled to learn “what it can and can’t do” as capabilities expanded
- •They decided to narrow focus to one high-performing feature
- 5:49 – 8:26
Idea #2 goes viral: AI that makes outbound calls for consumers
They discovered that users loved the voice calling feature: having the AI place phone calls on their behalf for tasks like reservations or appointments. Viral stories emerged, including negotiating car purchases across dealerships and long hold-time refund calls.
- •Voice calling became the standout feature users wanted most
- •Memorable use cases: negotiating with 10 dealerships, handling 1-hour hold times
- •Organic virality through influencers and mainstream attention (e.g., Variety mention)
- •The consumer app still exists but isn’t the company’s primary focus now
- 8:26 – 9:30
Why the viral consumer product didn’t monetize: novelty vs repeat usage
Zach explains the retention problem: people thought it was amazing but used it infrequently. Customer interviews revealed users liked it but wouldn’t pay for something they only needed a few times per year.
- •High “wow” factor but low repeat frequency (cars, refunds, rare tasks)
- •Users explicitly said they wouldn’t pay despite clear time/money savings
- •Team relied on direct outreach and interviews to understand behavior
- •They realized they needed a use case with ongoing, high-volume demand
- 9:30 – 11:27
The pivot trigger: inbound business demand—and a prank call that landed Omaha Steaks
Inbound requests from consumer users asked to deploy the voice technology for business phone lines. The Omaha Steaks CEO tried the consumer app, prank-called their COO, and immediately saw the call-center implications—becoming a major early customer.
- •Businesses asked to use the tech for inbound calls (not outbound)
- •Demo experience (calling the user’s own phone) made the realism obvious
- •Omaha Steaks deal originated from the CEO testing the app personally
- •They applied YC’s “make something people want” and said yes despite competition
- 11:27 – 14:01
Phone sales is mission-critical: Omaha Steaks’ holiday surge and staffing crisis
Cat describes why inbound phone sales matters for Omaha Steaks and similar DTC brands—especially during holidays. Omaha Steaks must scale headcount 15x with temporary workers, creating training/attendance issues and lost revenue from mishandled calls.
- •Many companies still drive significant revenue via phone, especially for complex/high-ticket purchases
- •Omaha Steaks is gifting-driven; demand spikes dramatically in Q4
- •15x seasonal staffing creates reliability and training failures
- •Poor call handling translates directly into large revenue loss
- 14:01 – 16:31
Deploying in the real world: integrating with 1990s-era systems and on-prem constraints
They share what it took to implement Simple AI at Omaha Steaks: two weeks on-site, understanding workflows, and integrating with legacy AS/400 terminal systems and slow-change IT processes. The integration effort became part of the company’s operational playbook.
- •On-site immersion with marketing, IT, leadership, and operations to map business reality
- •Legacy tech (AS/400 terminals, on-prem servers, limited deployment windows)
- •Integration depends on customer-built APIs and close collaboration
- •Customer commitment signaled by dedicating developers to Simple AI integration
- 16:31 – 20:26
From product catalogs to SKUs: teaching the agent to sell complex inventories
The agent must navigate messy, real-world product complexity: variants, bundles, promos, regional offers, and source-code driven campaigns. The founders describe translating this operational knowledge into an AI system that can confidently complete orders.
- •Needs customer context: new vs returning customer and purchase history
- •Order placement requires understanding variants, bundles, discounts, and seasonal promos
- •Marketing experimentation creates frequent catalog/promotion changes
- •They found that “every customer is complicated,” but patterns/abstractions exist across DTC sales
- 20:26 – 23:54
Performance edge: 30% better upsell than trained reps + personalization experiments
Cat claims the AI outperforms trained, full-time agents on upsells—key to unit economics in phone sales. They also describe new levers: changing voice/accent, tailoring to demographics and customer history, and rapid A/B testing via prompt updates.
- •Upsells drive significant revenue lift; small increases compound at scale
- •AI achieves 30% higher upsell rate than trained year-round human reps
- •Consistency: can ‘train on the best reps’ and apply best behavior to every call
- •A/B testing scripts and offers becomes fast and cheap (e.g., subscription wording that converts)
- •Personalization options include accent/voice selection and memory of past behavior
- 23:54 – 26:56
Beyond labor replacement: better customer experience through memory and conversation
They argue the AI can improve—not just cheapen—customer interactions by removing handle-time pressure and remembering personal preferences. Examples include customers sharing life stories and the system retaining preferences like not offering chicken again.
- •AI removes “average handle time” constraints that degrade human call-center experiences
- •Agent can be more conversational and attentive while still selling effectively
- •Persistent memory enables future calls to feel personalized and continuous
- •Customer notes can live in Simple AI’s database even when the client’s systems are limited
- •Positioning: focus on revenue lift and experience quality, not primarily cost savings
- 26:56 – 29:00
Company momentum: launch timing, funding, hiring, and building premium-quality voice AI
Cat and Zach share milestones: launching the current product about a year prior, growing from 2 to 10 people recently, opening a San Francisco office, and raising $14M. They emphasize a premium strategy—spending on latency and quality rather than racing to the bottom on cost.
- •Current version launched around February of the previous year
- •Team scaling and operational growth (office, fundraising)
- •Hiring plans: engineers, designer, and go-to-market roles
- •Premium approach: invest in R&D, natural voice, and fast response times
- •Founder-led sales transitioning toward a broader GTM function
- 29:00 – 35:18
Technical moat: latency, custom models, evaluation, and guardrails for reliability
Zach outlines the hardest engineering problems: responding in under ~600ms, building evaluation suites, fine-tuning models per customer, and solving long-tail issues like address transcription. They also stress orchestration/guardrails to prevent hallucinations and ensure correct order placement.
- •Latency is central to natural conversation; targets sub-600ms response
- •Evaluation infrastructure to simulate thousands of calls before production
- •Fine-tuning/RL and per-customer modeling for faster inference and better accuracy
- •Specialized components: address transcription, voice activity detection, end-of-turn detection per customer
- •Reliability requirements are stricter than typical chat agents (payments, orders, fulfillment)
- •Defensibility comes from deep production integration and solving many hidden edge cases
- 35:18 – 37:13
Founder lessons: don’t over-plan—talk to users early and follow demand
In closing, they reflect on iterating through three ideas and leaning heavily on user feedback. Their advice: start building, talk to users even when you only have a handful, and trust you can learn new domains if the problem is real.
- •Many founders wait too long to validate; they recommend shipping and talking to users immediately
- •Treat small user counts as seriously as large scale to learn faster
- •Domain expertise can be acquired—what matters is real pain and customer pull
- •Examples from YC lore (e.g., Brex pivot) reinforce following demand signals
