Jun 24, 2016 · 30m · 20vc
20VC: X.ai's Dennis Mortensen on Why There Is No Incumbency Advantage in Artificial Intelligence
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The 20 Minute VC, host Harry Stebbings interviews Dennis Mortensen, founder and CEO of x.ai, discussing the technical challenges of training AI scheduling agents, why big tech incumbents lack a data advantage, and x.ai's SaaS business strategy.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 26.9% of the talking time here. How this is scored →
Harry as informed peer 2.7 Guest teaching 5.3 Guest disagreement 3.7 Harry pushing back 1.5
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:39 Rejecting the big tech data advantage narrative
Dennis explicitly disagrees with the widely held VC consensus, stating that incumbents have zero useful data for interactive meeting agent negotiation.
Hardest push from Harry ▶ 21:47 Challenging pricing positioning via competitor comparisonHarry presses Dennis on market positioning by bringing up Clara Labs' $200 price point, questioning how x.ai can compete with drastically different pricing dynamics.
Biggest teaching moment ▶ 8:39 Educating on task-specific conversational dataset requirementsDennis educates Harry on why traditional databases like Gmail or Google Calendar cannot train dialogue models without active human-agent interaction data.
Harry holds his own ▶ 21:47 Demonstrating market awareness with competitor pricing dataHarry uses specific market data regarding Clara Labs' $200 monthly rate to push Dennis to defend his much lower software pricing model.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Dennis Mortensen's Background and Origin of x.ai | 2 | 7 | 3 | 1 | Harry asks open setup questions about Dennis's background and x.ai's technology. Dennis provides an extensive breakdown of data labeling complexities and the hidden grunt work required to train AI agents, correcting popular misconceptions about rapid AI development. | |
| Data Advantage and Incumbency in AI | 3 | 8 | 7 | 2 | Harry brings up the common VC perspective that tech giants hold an unassailable data advantage. Dennis strongly pushes back against this premise, explaining why passive email and calendar data are useless for task-oriented agent negotiations. | |
| Enterprise Productivity and Democratizing Virtual Assistants | 2 | 5 | 2 | 1 | Harry conveys questions from investors regarding enterprise integration and vision. Dennis outlines how AI agents will democratize assistant capabilities across corporate organizations, while Harry largely agrees and validates his perspective. | |
| Fundraising Strategy and SaaS Monetization Model | 3 | 4 | 2 | 2 | Harry inquires about x.ai's tactical fundraising journey and pricing strategy. Dennis breaks down his logic for selling an answer rather than product traction during early rounds and explains his planned SaaS tier pricing. | |
| Competitive Landscape and AI Network Effects | 4 | 6 | 5 | 2 | Harry challenges Dennis by citing high competitor pricing models such as Clara Labs ($200/mo). Dennis rejects the comparison, clarifying the economic difference between human-in-the-loop services and scalable AI software with network effects. | |
| Quick Fire Round with Dennis Mortensen | 2 | 2 | 3 | 1 | A friendly quick-fire round covering books, location preference, and fun user interactions. Dennis offers a mild contrarian view regarding mainstream pessimism around AI existential risk. |
Statements from this episode (14)
Stebbings: x.ai has raised over $30M in venture capital
“And x.ai is now one of the best funded AI startups, having raised over thirty million dollars from our friends at Firstmark, and a big thanks to Matt Turt for making the intro, and the likes of DCM, Lira Hippo, Crunch Fund, SoftBank, and many more incredible i…”
Mortensen: x.ai spent 1.5 years building its time-parsing AI model
“That alone, we probably spent a year and a half coming up with the right model, because humans will say stuff like, we'll meet up later, this afternoon.”
Mortensen: x.ai wrote 26-page guidelines just to label single data entities
“Just one of those entities, which I mentioned, whether it be time or people location, that's a 26 page set of guidelines just to label it.”
Mortensen: AI agents require massive data grunt work, not quick TensorFlow hacks
“Anybody who believes that it's about getting five good engineers into the basement and we can start hacking away on TensorFlow come Sunday is probably going to be quite surprised.”
Mortensen: Tech incumbents have zero usable data for training AI scheduling agents
“There you have no data set, as in zero, nothing, because there is no agent-human negotiation under a given dialogue model that exists in that data set, and you can't label it in the past, as in, what if the agent said this, what would the human then say?”
Mortensen: Enterprises will buy third-party AI agents rather than build them
“And I think it's going to be equally silly to go, as an enterprise, say, I want to engineer ERP. A meeting scheduling agent like Amy, an invoicing agent like Tommy, a some agent like this. I think they'll have to go into the market and figure out which of thes…”
Mortensen: x.ai will derive most of its revenue from enterprises
“I think in the end, X.ai as a company will drive most of its revenue From the enterprise.”
Mortensen: AI agents will not cause unemployment among human assistants
“This is not about making sure that any human personal assistant becomes unemployed in the future. I think all of them will continue to be employed, and I think they'll thrive and do proper work, not email ping pong.”
Mortensen: x.ai's $23M Series B provides runway deep into 2018
“And if we did nothing and just kept everything intact, we will be way into 2018. On making no revenue existing monthly investments.”
Mortensen: x.ai will launch with free, pro, and business pricing tiers
“I'm very confident you'll see us be in market with three editions. A free edition where you can schedule a few meetings in the name of Amy or Andrew at x.ai. A professional edition for where you can schedule unlimited meetings, tune the signature, and do some …”
Mortensen: Human-in-the-loop assistant services top out as $30M businesses
“Sell them between two to 500 dollars, which is what they charge, which is not necessarily expensive compared to having a full-time human assistant in your office, but it's still an outsourcing play of sorts, and I'm not saying that to dismiss them, it's just a…”
Mortensen: AI scheduling assistant market will be winner-takes-all
“Whoever solves the software challenge will have this opportunity to take the market. It's a potential winner takes all, or certainly winner takes most, given the positive outcome of Amy speaking to Amy.”