What is one specific practice or strategic choice that has most enabled rapid revenue growth in your AI company or portfolio, and why?
Most answers emphasize leveraging proprietary data, rapid iteration, or specialization to drive growth, though some focus on renting core models, maximizing usage, or building developer-focused products.
10 answers
Answers


Reid Hoffman
Co-Founder, Executive Board Chair · Manas Ai
Close feedback loops quickly
Reid Hoffman says a founder's job is to constantly close as many feedback loops as possible as quickly as possible, then add leverage, describing an AI-native company as a set of feedback loops hill-climbing against an objective function.
22nd Century Frontier® · 22nd Century Frontier® · Read

David Haber
Board Member · Partnership For New York City
Rent Core AI Models
David Haber says companies should rent core AI models from the best provider rather than build them, because whatever they build themselves will get commoditized in six months as frontier models advance.
Fintech Leaders · Fintech Leaders · Read

Ben Horowitz
Horowitz · Andreessen Horowitz
Build a developer stack
Ben Horowitz cites ElevenLabs getting into the developer stack and building the better product for developers, which gave it built-in customers and a data feedback loop that improved its model.
The answer starts at 41:47.
a16z Podcast · Listen on Apple Podcasts

Michael Mignano
Partner · Lightspeed Venture Partners
Build products quickly and cheaply
Michael Mignano says it is getting dramatically easier to build a product, noting his firm incubated a product built very quickly and cheaply, though he frames this as raising the bar for standing out rather than as a revenue driver.
The answer starts at 17:07.
Sourcery · Listen on Apple Podcasts

Daniel Ives
Partner and Senior Managing Director, Tech Analyst
Maximize token consumption
Daniel Ives says his belief is that token consumption leads revenue, so if token consumption ramps aggressively, revenue follows.
The answer starts at 6:55.
The Pomp Podcast · Listen on Apple Podcasts

Peter H. Diamandis
Founder & Curator · Abundance360
Proprietary Domain Data
Peter H. Diamandis says entrepreneurs who have specific use-case data can build a company by using it to tune or train a model, because such specialist models have a good chance of outperforming foundation lab models in that use case.
The answer starts at 14:01.
Moonshots with Peter Diamandis · Listen on Apple Podcasts
Ben Gaffney
Deputy General Counsel · Openai
Proprietary data
Ben Gaffney says having proprietary data is one of the things that sets a company up for success, because it is much harder to replicate than what the technology can do by itself.
Shift AI · Shift AI · Listen at 53:18

Anish Acharya
General Partner · Andreessen Horowitz
Specialize deeply, charge high prices
Anish Acharya says narrow startups should build small, go deep and charge a lot, arguing that specialization is a new moat because collapsing software costs let a company go so much deeper for one customer that competitors would need years of roadmap to match it.
EO · EO · Listen at 8:19

Sunil Dhaliwal
Founder and General Partner · Amplify Partners
Invest in strong technical founders
Dhaliwal says his firm's through line is backing technical founders with insight into emerging problems.
N/A · N/A · Listen at 14:18

David Weisburd
Co-Founder · Weisburd Pierce
Build AI agents
David Weisburd says that starting from scratch today he would build AI agents to handle different parts of the venture capitalist role before hiring any humans, and would make everything legible to AI and LLMs to create a learning loop for the agents.
How I Invest with David Weisburd · How I Invest with David Weisburd · Listen at 35:32
More Generative AI questions
View all →- What design choice improves AI agent privacy?5 answers
- What architectural choices enable reliable agentic AI systems?8 answers
- What responsibilities should AI creators have for their models?12 answers
- How does your team create feedback loops for AI models?12 answers
- What architecture optimizes compute for AI workloads?13 answers


















