Mar 12, 2026 · 44m · product-market-fit
This 3x founder hit $1M ARR in 5 months. Here's his playbook. | Roy Moussa, Founder of GetVocal · PMF Show
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 'The Product Market Fit Show,' GetVocal founder Roy Moussa details how his third startup scaled to $1M ARR in five months by pioneering context graph architecture for enterprise AI customer service. He shares key technical innovations, product-market fit strategies, and actionable advice for building resilient, high-growth enterprise companies.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Pablo holds 16.1% of the talking time here. How this is scored →
Pablo as informed peer 4.0 Guest teaching 4.5 Guest disagreement 1.9 Pablo pushing back 1.9
speaking balance: gold is Pablo, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 18:57 Roy dismisses chat-only enterprise AI as useless
Roy takes a hard, critical stance against conventional conversational bots, asserting that AI that merely chats without executing deterministic processes provides zero enterprise value.
Hardest push from Pablo ▶ 36:00 Pablo pushes back on effortless enterprise agent deploymentPablo challenges Roy's optimistic framing of rapid self-serve agent sprawl at Glovo, pressing him on the actual engineering overhead and operational friction required.
Biggest teaching moment ▶ 25:46 Roy educates on Speech Act Theory and intention networksRoy breaks down the fundamental difference between standard semantic knowledge graphs and intention-driven context graphs based on linguistic Speech Act Theory.
Pablo holds their own ▶ 22:15 Pablo articulates the core deterministic vs probabilistic trade-offPablo demonstrates sharp technical intuition by precisely describing how rigid deterministic chatbots get trapped in edge cases while pure probabilistic LLMs introduce unacceptable business risks.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Pablo as informed peer | Guest teaching | Guest disagreement | Pablo pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome and Roy Moussa's Entrepreneurial Background | 3 | 3 | 1 | 1 | Pablo welcomes Roy and frames his rapid European fundraising success before prompting him on his engineering roots. Roy collegially details his background across aerospace and early AI startups. | |
| Roy's First Startup and Lessons from Computer Vision | 3 | 3 | 1 | 1 | Pablo asks Roy about the size and scope of his prior ventures. Roy explains his 2013 horizontal machine learning platform and its eventual pivot into computer vision for retail logistics before exiting. | |
| Pivoting to Conversational AI and Founding GetVocal | 2 | 3 | 1 | 1 | Roy shares the origin story of GetVocal, starting with an Alzheimer's conversational agent before moving into enterprise sales workflows alongside their former VP of Sales. | |
| Evolving GetVocal's Value Proposition and Focus | 3 | 4 | 2 | 1 | Roy explains how GetVocal expanded from narrow customer support into an interconnected enterprise agent fleet, highlighting the distinction between horizontal platforms and vertical wedge adoption. | |
| Testing Early Sales Automation Use Cases | 4 | 4 | 2 | 3 | Pablo drills into the early sales automation use cases, pressing Roy on whether experiments were run serially or in parallel and how they determined urgency across customer ICPs. | |
| The First Breakthrough Moment in Customer Support | 4 | 4 | 2 | 3 | Roy describes a breakthrough weekend with a telecom client where resolution metrics spiked. Pablo pushes to clarify whether the agent was merely conversing or executing actions. | |
| Why Traditional Chatbots Fail in Enterprise CX | 4 | 6 | 4 | 2 | Roy forcefully dismisses chat-only AI as useless and critiques traditional chatbots and rigid architectures, citing high-profile industry failures like Sierra at Gap. | |
| Mid-Episode Listener Callout | 6 | 5 | 2 | 2 | Following a quick callout, Pablo articulates the exact architectural dilemma between brittle deterministic decision trees and unpredictable probabilistic LLMs, which Roy enthusiastically confirms. | |
| Technical Deep Dive: Context Graphs and Intention Networks | 4 | 7 | 2 | 2 | Roy takes Pablo through a deep technical breakdown of context graphs and Speech Act Theory, explaining how intention networks provide deterministic memory and auditable reasoning. | |
| Ingesting Enterprise Knowledge & Building Context Graphs | 5 | 6 | 1 | 2 | Pablo asks how context graphs handle messy enterprise documentation and edge-case policies. Roy outlines how conflicting sources are resolved and codified into deterministic nodes. | |
| Measuring Agent Performance and Product Feedback Loops | 5 | 5 | 2 | 2 | Roy explains syllable-level tracking and product feedback loops. Pablo defends diving into granular technical product mechanics to understand true defensibility against competitors. | |
| Organic Virality and Fleet Expansion at Glovo | 5 | 5 | 2 | 3 | Roy shares Glovo expanding to 80 agents in 8 weeks. Pablo challenges the notion of effortless enterprise expansion, asking how much manual forward-deployed engineering is actually required. | |
| Competitive Landscape and Human-in-the-Loop AI | 5 | 5 | 3 | 3 | Pablo introduces competitor Decagon to compare positioning. Roy contrasts their architectural approach and underscores the necessity of human-in-the-loop systems for enterprise reliability. | |
| Reaching $1M ARR in 5 Months & Founder Advice | 3 | 3 | 1 | 1 | Roy shares hitting $1M ARR in five months with a single salesperson and offers closing advice on founder agility and customer focus. |
Statements from this episode (17)
GetVocal raised a $3M Seed and $26M Series A in one year
“Last year, you know, you raised a three million dollar seed round at the beginning of the year. By the end of the year, you'd raised like a twenty six million dollar Series A.”
Coding is becoming less relevant for tech founders
“Not to mention that coding is less and less a thing”
Customer service currently sees the highest enterprise AI buying adoption
“We narrowed in obviously on the use case that is in most transformation, which is customer service particularly. And that's where Cost pressures are high, and that's where, you know, adoption or buying at least is there.”
GetVocal originally automated SDR workflows before pivoting to customer support
“We went in and put commercial agents, PDR agents, actually in the beginning, doing a lot of the heavy lifting on those, from that, those processes”
AI automation delivers instantaneous ROI in headcount-heavy customer support teams
“I was definitely more enterprise and further down the customer journey where it was very FTE heavy. It was very people heavy organizations, internal or externalized, and that the ROI was instantaneous.”
Conversational AI that only chats is useless for enterprise businesses
“For businesses, AI that just chats is useless.”
Most AI chatbots break beyond 10% to 20% of simple cases
“Most chatbots and AI agents out there hit a ceiling and break
Past the 10% simple cases. Okay, call it 20% depending on the organization.”
Sierra suffered a massive AI agent failure with Gap
“There was a massive hiccup with AI agents that Sierra launched for Gap.”
Purely probabilistic AI makes no sense for deterministic business processes
“Business conversations are deterministic, and it makes no sense to put a pure probabilistic thing to do a mostly deterministic process.”
GetVocal based its context graph architecture on Speech Act Theory
“Three years ago, we took that, and we thought, okay, we built from the ground up, purposely designed for processes and conversations, and we based it on something called the Speech Act Theory, a new class of graphs.”
Structuring graphs around user intention is best for conversational memory
“That actually interconnects things, not based on semantic, which is knowledge, but based on intention, which is in any conversation, in any process, there's one single unique intention at a time. And so it's a network of interconnected intentions. Anyways, we …”
Standard RAG fails in enterprises due to conflicting internal knowledge
“Just a standard LLM doing rag onto these knowledge sources and trying to answer questions is often you have conflicting answers from two different sources because that's the reality of enterprise building up knowledge bases across You know, a huge company.”
Repeat customer service processes should be deterministic, not probabilistic
“Whenever a pattern is detected, a process that is defined, it's being repeated. We're servicing that case every time in the same way. Why keep it probabilistic and call upon a model every time? It has become deterministic.”
GetVocal doubled vendor uptime for Glovo restaurants and grocery stores
“And we deployed for one use case, showed a lot of value, a lot of ROI. We doubled their uptime in the restaurants they serve and groceries and a bunch of other KPIs like that.”
Expecting AI agents to perform immediately without human coaching is unrealistic
“Expecting something from the get go to be the highest performing thing is like getting somebody from the street, giving them a piece of paper for instructions and telling them, okay, now you're doing this job. Good luck. You have to perform by tomorrow.”