Ignite AI: Dennis Mortensen on Startup Failure, AI Agents, and Why Boring SaaS Problems Win | Ep268

What does a founder learn after selling four companies, burning one to the ground, and spending years building AI agents before “AI agents” became the phrase of the moment?

Dennis Mortensen has the scars to answer that. A Danish-born, New York-based serial founder, Dennis has built and exited companies across analytics, media optimization, and AI, including X.ai, the AI scheduling assistant that raised $44 million from FirstMark and others before being acquired by Bizzabo in 2021. Today, he is building LaunchBrightly, a company automating product screenshots for help centers—a problem that sounds painfully unsexy until you realize every software company has it, every product team pays for it, and every outdated screenshot quietly creates support debt.

In this episode, Dennis joins Brian Bell for a wide-ranging masterclass on founder judgment, painful pivots, and the unglamorous infrastructure problems that make or break software companies.

In Today's Episode We Discuss:

00:01 – Meet Dennis Mortensen

01:25 – From IBM Dreams to Serial Founder

03:51 – Selling His First Company During the Dot-Com Era

05:00 – Building in Budapest and Moving to New Yor

07:08 – Why European Founders Look West

09:25 – The “Expensive MBA” Startup Failure

11:52 – Why Dramatic Pivots Are Overrated

13:51 – The Marketplace Mistake That Killed the Business

16:27 – When the Market Is Telling You You’re Wrong

18:23 – The Twitter Pivot and Founder Mythology

20:46 – Why Business Model Flexibility Matters

22:35 – Founder Bias, Persistence, and Not Dying

24:30 – Shutting Down and Moving On

26:34 – Building IndexTools and Real-Time Analytics

31:34 – Why Founders Should Take M&A Calls

35:05 – How Optionality Creates Future Exits

36:45 – From Yahoo to Visual Revenue

40:01 – The “List of Hate” Startup Ideation Process

44:57 – Why Founder Focus Beats Angel Investing

48:43 – Building Visual Revenue for Digital Publishers

53:44 – Selling Visual Revenue to Outbrain

54:58 – The Pain Behind X.ai

55:26 – Market Challenge vs. Science Challenge

56:59 – Why Scheduling Was a Worthy AI Problem

01:00:40 – Testing X.ai with Human Assistants First

01:02:31 – Wizard-of-Oz Testing and Scheduling Complexity

01:05:34 – Building AI Before Modern LLMs

01:06:07 – 47 Intents and 32 Million Labeled Data Points

01:10:13 – Lessons from the X.ai Journey

01:11:14 – Why Winning the Turing Test Was the Wrong Goal

01:14:55 – When Customers Stop Being Sold and Start Buying

01:17:04 – Introducing LaunchBrightly

01:17:43 – Building for the Love of the Sport

01:19:20 – Why LaunchBrightly Exists

Pull quotes:

“If you can just figure out a way to just not die, that’s probably the best way you can somehow win.”

“I have a little list of hate on my phone.”

“It’s okay to be a machine doing machine things in a machine-like way.”

Dennis’s story is not the sanitized founder mythology of perfect timing and clean wins. It is a sharper, more useful version: build, sell, fail, learn, repeat—and keep choosing problems painful enough that someone already has a human doing the work.