Dharmesh Shah on Building Businesses With AI

TL;DR

Generative AI represents a broad business opportunity because it can affect nearly every industry and workflow, according to HubSpot co-founder Dharmesh Shah. Entrepreneurs should actively experiment with tools such as ChatGPT, combine models into simple workflows, and look for practical products built on top of large language models, while recognizing that developing the underlying models requires substantial capital.

Transcript

again I've been in software for 30 years now doing startups pretty much my entire professional career the only time I've felt like like how hard palpitations kind of like Sean kind of open with us like there's this party going on next door and I'm here knitting right it's like this is like too big to ignore I... Read More

Key Insights

• Generative AI is described by Dharmesh Shah as the largest technology opportunity and paradigm shift he has encountered during 30 years of building software startups, because its potential applications extend across businesses rather than remaining limited to a narrow set of consumer use cases.
• Mobile technology created important but discrete opportunities by adding capabilities such as cameras and GPS to phones, enabling consumer applications such as Uber. Shah contrasts that pattern with generative AI, which he believes can affect nearly everything in a manner closer to the original internet shift.
• OpenAI is presented as a leading candidate to become an exceptionally valuable private company because its technology may be highly transformative. The participants place it among the most plausible companies for enormous future valuation, although Shah does not identify definitive second or third candidates.
• Large language model development is highly capital intensive, which is unusual for a software company. Shah argues that OpenAI needed access to considerable funding and structured its for-profit operation with capped profits as a practical response to the expense of building foundational models.
• OpenAI's move toward a for-profit subsidiary is framed as a funding solution rather than clear evidence of deception. The discussion recounts a reported story in which Elon Musk withdrew intended support after a rejected takeover proposal, leaving the organization with a major funding shortfall.
• Hands-on experimentation is presented as an effective way to understand generative AI. Shaan Puri cleared his calendar for an AI-focused week because he felt the opportunity was too significant to observe passively, then explored tools by building small projects and testing combinations himself.
• AI tools can be combined into useful creative workflows without requiring users to write code. Puri generated an original podcast rap with ChatGPT, recorded himself performing it, and processed the recording through a Google Colab interface connected to a community-uploaded voice model.
• Synthetic voice technology can reproduce a recognizable performer from an ordinary user recording. The demonstrated workflow reportedly took about 15 minutes, showing that capabilities once demanding specialized production expertise can be accessed through shared models, hosted files, and simple notebook interfaces.

Questions & Answers

Q: Why does Dharmesh Shah think AI is bigger than mobile?

Dharmesh Shah argues that mobile created a meaningful but relatively discrete set of opportunities. Adding cameras and GPS to phones enabled consumer products such as Uber and produced many successful startups. Generative AI appears broader to him because it can affect nearly every kind of work and business. He compares that reach to the arrival of the internet and calls AI an order of magnitude larger than mobile.

Q: How should entrepreneurs begin exploring generative AI?

Entrepreneurs can begin by reserving focused time for direct experimentation instead of only reading about the technology. Shaan Puri cleared his calendar for an AI-focused week and tried different tools because hands-on tinkering is how he learns. His example suggests starting with a small, personally useful creative project, testing available interfaces, and combining tools until a working process emerges.

Q: How can ChatGPT help generate a creative business asset?

ChatGPT can turn a short brand prompt into usable creative material. Puri asked it to write an introductory rap for the My First Million podcast and supplied the show's key phrase, "no small boy stuff." ChatGPT produced a complete rap with an on-brand chorus about ambition, money, and progress. He then used that text as the input for a separate synthetic-voice experiment.

Q: How was an AI-generated podcast rap created?

The process combined several existing AI resources. First, Puri asked ChatGPT to write an introductory rap based on the podcast and its key phrase. He then recorded himself performing the generated lyrics. Finally, he used a Google Colab workflow linked to a community-uploaded Kanye voice model, which transformed his vocal performance into a synthetic version resembling Kanye's voice.

Q: What makes large language model companies capital intensive?

Dharmesh Shah says building large language models at OpenAI's scale is supremely capital intensive, an unusual characteristic for a software company. That expense creates a need for access to substantial outside capital. In the discussion, this funding requirement helps explain why OpenAI established a for-profit operation and raised investment rather than continuing to rely entirely on the resources available to a nonprofit research organization.

Q: Why did OpenAI create a for-profit subsidiary?

The account presented in the conversation says OpenAI was spending heavily on research and needed more money. After Elon Musk reportedly withdrew planned financial support following a rejected proposal to take control, the organization faced a large funding shortfall. A for-profit subsidiary allowed it to raise investment capital, while a capped-profit structure placed limits on the returns generated by that entity.

Q: What was Elon Musk's reported disagreement with OpenAI?

The hosts recount a reported story that Elon Musk believed Google was far ahead and proposed taking over OpenAI to redirect it. Sam Altman, Greg, and the group running OpenAI reportedly rejected that plan. Musk then withdrew his funding and left the organization. The participants emphasize that they lack insider knowledge, and they present this narrative as a published report rather than a confirmed firsthand account.

Q: Why are simple AI tool combinations a business opportunity?

Simple combinations can make sophisticated capabilities accessible without requiring customers to build foundational models or write code. Puri's experiment joined ChatGPT, a personal voice recording, a Google Colab interface, a hosted file, and a community voice model. The process reportedly took about 15 minutes. That example suggests entrepreneurs can create practical products by organizing existing AI components into convenient workflows for specific outcomes.

Summary & Key Takeaways

• Dharmesh Shah describes generative AI as the largest technology shift he has experienced during 30 years in software and startups. Unlike mobile, which created a more discrete collection of camera, GPS, and consumer-app opportunities, he believes AI can influence nearly every business and become an opportunity comparable to, or larger than, the internet.

• The discussion examines OpenAI's organizational and financial evolution. Large language models require unusually high amounts of capital for a software business, so OpenAI needed substantial funding. The hosts recount a reported disagreement involving Elon Musk and explain how a for-profit subsidiary with capped profits offered a way to attract investment while retaining the original mission.

• Shaan Puri demonstrates how quickly entrepreneurs can explore AI by combining existing tools. He used ChatGPT to write an on-brand podcast rap, then applied a community-uploaded Kanye voice model through a Google Colab workflow. The experiment took about 15 minutes and illustrates how accessible tools can be assembled into new creative products without writing code.