A Conversation with Jeff Bussgang on "The Experimentation Machine"
Alvaro: Good afternoon, and welcome to this Far Side Chat with Jeff Bussgang. Jeff is truly a leader in entrepreneurship and venture capital. We've known each other for over 30 years as HBS classmates and later as colleagues. Jeff co-founded Flybridge over 20 years ago, which has grown into a premier early-stage VC fund with over a billion dollars under management. He's written three books and has been on Harvard Business School's faculty for over a decade, leading the popular "Launching Tech Ventures" course that one-third of MBA students take.
His new book, "The Experimentation Machine," isn't just about becoming a 10x founder – it's about becoming a 10x leader with AI. It's both an invitation and a practical manual for startup founders, with step-by-step guidance on AI implementation.
Jeff: Thanks so much, Alvaro. Yes, we've been friends for what feels like forever! It's been a joy to share these parallel journeys as investors, entrepreneurs, and faculty colleagues.
Speaking of AI adoption, I actually revealed at the book launch that my 92-year-old mom is a power user of ChatGPT! She was telling some of my students how she used it to write a job description for the choir director at her senior care home. So really, if my mom can master these tools, anyone can.
Alvaro: That's amazing about your mom! Your book demystifies AI. You mention that a PhD is no longer needed to build an AI-forward company because the technical playing field is being leveled.
Jeff: Absolutely. The core message of the book is that AI is not going to replace founders anytime soon, but founders who use AI will replace founders who don't. Those who learn to use AI effectively will become "10x founders" – ten times more productive than the average founder. And this applies beyond founders to salespeople, analysts, consultants, marketers, and even parents.
I've been an entrepreneur and investor through multiple tech waves – web 1.0, cloud, mobile, and crypto. But I firmly believe this moment is far more impactful and disruptive than any previous technology wave. It will affect all 8 billion people on this planet, and it's happening incredibly fast.
Alvaro: Can you explain why you think AI adoption is different from previous technological revolutions?
Jeff: Unlike previous technologies that followed gradual adoption curves, AI adoption is "absolutely vertical" in its speed. One key reason is that the delivery devices already exist in everyone's pockets – 8 billion smartphones, over 2 billion laptops and tablets, and 1.5 billion automobiles that can receive over-the-air updates.
Think about my first internet transaction in 1994 when we were students – I "foolishly" put my credit card into a website to buy a mug from a small e-commerce store. It took decades for e-commerce to become universally available through Amazon and Shopify. With AI, the capabilities are being delivered immediately to existing devices.
Look at what's happening with major companies – Google recently announced that one in four lines of code they write is constructed by AI. Andy Jassy at Amazon reported that their AI-infused platform has saved them 4,500 developer years – in his words, "yes, that number is crazy but real."
Alvaro: Let's talk about the concept behind your book title. For those unfamiliar with the term, what exactly is an "experimentation machine"?
Jeff: It's a great question. The concept has roots in scientific discovery, where scientists realized there was a whole world of knowledge that could be discovered through experiments. You form a hypothesis, design an experiment to test it, analyze results, and iterate.
Eric Ries with "The Lean Startup" and Steve Blank with "Four Steps to the Epiphany" brought this mindset to entrepreneurship. Instead of spending years building a product, you create a minimum viable product in weeks or months, see how customers react, and iterate.
In early-stage entrepreneurship, you want to maximize learning, not metrics. AI supercharges this approach by letting you run, design, and execute more experiments than ever before. The key is using strategic reasoning to sequence experiments that reveal the most valuable outcomes.
Alvaro: I love how you bring a scientific approach to entrepreneurship. A startup is fundamentally a learning machine where you form hypotheses and run experiments to test them.
Jeff: Exactly! Let me give you an example. Booking.com, which most people know, prides itself on running thousands of experiments daily – testing different value propositions, language, imagery, and features – to optimize results. We backed a team from Booking.com a few years ago, and their execution speed was unlike anything we'd seen before.
It's not just nice words on a page – it's a mindset, almost a religion. Instead of debating "should we go left or right," the experimentation mindset says "I don't know and I don't really care – let's design an experiment, test it, look at the data, make a decision, and move on." That's the beauty of this approach.
Alvaro: Absolutely. This mindset isn't just for startups – any organization can gain a tremendous edge by adopting this experimental approach to learning and decision-making.
The Experimentation Mindset: AI as a Game-Changer for Businesses
Alvaro: In your book, Jeff, you have an interesting quote that says, "anyone approaching building and running a startup as they would a normal business is doomed to fail." What I think is behind that quote is that anyone approaching a startup without an experimentation mindset is headed for trouble. I'd take it even further—even large corporations today and tomorrow need an experimentation mindset because it's not about who's right, it's about running experiments so the data can tell us what's right.
Jeff: Absolutely.
Alvaro: Now, let's get into AI. Tell us how this experimentation mindset can actually benefit from introducing AI into these processes.
Jeff: What I see our best founders doing—and we now as a firm only invest in AI companies, which has been the case for several years—is using these tools throughout every step of the ideation process, customer discovery process, go-to-market process, and discovering a profitable business model.
Our colleague Tom Eisenman created this diamond square framework, which goes through each component of the business model, and our students and founders are applying AI tools to every single component, every step of the way. They're even doing it before they get started—we talk about the zero-to-one process, but many students and founders are going from the negative one to zero process using AI tools.
They're using AI for research and customer discovery, transcribing customer interviews and recording them into tools like Google's notebook LM, which allows you to put knowledge into a system and train chatbots against that knowledge. They're creating customer personas using AI so their team can have 24/7 conversations with these customer models to test different value propositions and features. And they're certainly building amazing go-to-market strategies to enhance their sales team's capacity.
Alvaro: I've become a huge fan of the diamond square created by Tom, and I love how you put it in the book—what every startup needs is to build a product, acquire customers, satisfy customers, and make money. Can you give us some examples of how companies are using AI to be much more effective in go-to-market strategies?
Jeff: Sure. Many of our portfolio companies' sales organizations now use AI before reaching out to prospects. If you're approaching a public company, there are quarterly earnings calls, press releases, competitive analysis, trade show information, sponsored content—a whole bunch of content out there. There are tools now that draw from that content to enrich your data set for prospecting, qualify and score prospects, and brief your sales team before they walk into a sales call.
Then, some tools allow you to generate personalized demos for each prospect. Some of our portfolio companies, like Topline Pro, have created thousands of personalized demos with an AI avatar representing a sales representative who speaks directly to the customer, shows them a product with their content, and lets the prospect get a sense of what the quality of work will look like.
We're seeing this even in mature companies. I have a portfolio company that had 1,500 employees in 2023 and now has 1,100 employees, but has grown revenue by 50%. They've effectively automated much of their customer service team, improved sales team efficiency, and increased their software development velocity with code generation tools. It's not just early-stage startups doing this—scale companies are implementing these tools effectively as well.
Alvaro: Most of our audience here are from emerging markets. I'm sure many are thinking this all sounds great for US companies or US startups. When you talk about understanding customers better—asking questions like "what does a day in the customer's life look like?" or "what tasks do they perform?" or "where do they experience friction?"—if you're an emerging markets company, will you get relevant information? Do these tools work outside the US?
Jeff: Absolutely. It's not only that the models are trained on specific countries and use cases—they're also multilingual and multifaceted. If there's a data set you feel is missing, you can add your data to the model to improve its knowledge base.
I mentioned notebook LM, and there are also custom GPTs from OpenAI and artifacts from Claude and Anthropic. If you have specific customer data that only you know—nuances and earned secrets built over the years—you can feed that into these models securely and then derive insights, all in Spanish, Portuguese, English, Japanese, or whatever language you prefer. It's truly a global tool, which makes it so exciting.
Those in the audience should just try playing with ChatGPT's voice functionality—ask questions in whatever language you want, have it read documents in any language. You'll be amazed at its capabilities. We all need to start experimenting and interacting with these LLMs daily to become AI-native. Eventually, it becomes second nature as a thought companion.
Alvaro: Let's talk about speed. In your book, you use the term "moving faster with AI." Tell us why and how you can obtain not just productivity but speed.
Jeff: It's a great question because this isn't just an efficiency play—it's a capabilities expansion. I remember in the early days of the internet, we used to joke about "internet time" being like dog years, where every year in 1996-98 felt like seven regular years. Today, every month in the AI world feels like seven years!
The speed of execution we're seeing from the best companies indicates what all companies can achieve using these tools. The challenge is that humans have to adjust their operating model to work at this speed because usually, we're the points of friction. We're habitual creatures—Excel just reached its 30-year anniversary and remains incredibly popular because we're all used to it.
We need to build organizations that adapt and execute faster. As you noted earlier, this requires a willingness to experiment and then make quick decisions to scale those experiments when they succeed. It's not just about testing a hypothesis off to the side—it's about shifting the entire organization to a new way of doing things if the hypothesis proves out. That's the change management challenge for leaders in the coming months and years.
Alvaro: When we talk about AI, the human side inevitably comes up. As one of your students quoted in your book asks, "Why even have human founders?" What's the human element in all this? Are we going to get rid of people because these agents are becoming so smart and making decisions for us?
Jeff: I think the human element is going to center around strategy, insights into markets, leadership, taste, discernment, and curation. These will matter more and more as rote tasks get automated away.
We've seen this pattern before—fewer bank tellers with ATMs, potentially fewer taxi drivers with self-driving cars, and reduced agricultural workforces from farm automation. But I don't think humans or work will disappear—we'll just be more efficient.
The economist John Maynard Keynes, observing the industrial revolution, predicted his grandchildren would work just 20 hours a week. Well, I don't know about you, Alvaro, but I don't work 20 hours a week, and nobody I know does! We're all working hard despite all the automation tools we've benefited from over recent decades. I think AI will prove similar—we'll just do more things, be more efficient, and develop more capabilities. We as humans will be better at leveraging these incredible new tools.
Alvaro: You say in your book that we're still in the early innings of the generative AI revolution. If this is evolving so fast, why should I get into it now rather than waiting until it's more mature?
Jeff: Because your competitors are getting into it now and learning how to use these tools. I'm a bit self-conscious writing an AI book given how fast it's all moving—in five years, the specific tools I describe may not still be around, but the mindset of leveraging and using the tools and building that muscle will remain.
My advice is to build the muscle now, don't wait. It's like when a trainer tells you to start preparing for a marathon a year in advance. You can't just train right before the race—you have to build up that capability, that stamina time and time again, and be prepared when the race begins.
Alvaro: Agreed, never too early. There are a lot of skeptics about AI for many reasons. What do you say to the AI skeptics?
Jeff: There are usually two types of skeptical points. One is that it's overhyped—all buzz and froth with no real substance. But I can tell you the substance is in your phone, on your laptop, and in your car. The AI capabilities are here and being used now.
Look at the "Magnificent Seven"—the seven most valuable companies in the world, like Google, Amazon, and Facebook. These are the smartest, most sophisticated companies, growing revenue 20-30% year-over-year while keeping headcount flat, becoming more efficient and growing in valuation by leveraging AI. So, to the first point: look at the underlying metrics and productivity—it's real.
The second concern is that AI will result in greater inequality—a greater gap between haves and have-nots, both by class and geography. Will only the US and maybe China have these capabilities and exploit them? These are fair risks, but all the research shows these capabilities are having a massive impact on the bottom of the pyramid.
One research study by our colleague Professor Ram Cunning from Harvard Business School looked at Kenyan entrepreneurs using generative AI in their small businesses. They dramatically improved profitability and performance even as relatively unsophisticated users in a developing country. The research is clear—this technology has an enormous impact on emerging markets and entrepreneurs who are leapfrogging technology waves to get right to this one.
Alvaro: It's time for us to get into Q&A. One question relates to investment opportunities in this sector, with a tone of "Have I missed the boat? Is it too late to get into this?" Where do you see opportunities from your perspective?
Jeff: I have a skewed lens as a venture capitalist at Flybridge. We invest in early-stage AI companies, and we've never been more enthusiastic. We think the current cohort of companies we're investing in is as promising as ever.
If you look back 12-13 years ago, Aileen Lee coined the term "unicorn club" and profiled the 20-30 companies valued at over $1 billion. Today, the unicorn club numbers in the thousands, and we talk about "decacorns" ($10B companies) and "centacorns" ($100B companies). We even have a handful of trillion-dollar companies, which were unimaginable 10-20 years ago.
So the provocative question is: Do you think there will be fewer $100B companies or $10B companies created in the next 10 years compared to the last 10? Do you think the AI revolution is played out? I don't think so. I think we're in the second or third inning of this game, or maybe 5-10 minutes into a 90-minute football match. We have a long way to go, and a huge amount of value will be created, not only disrupting the IT industry but also eating into the multi-trillion-dollar staffing industry behind it.
Alvaro: There are a couple of questions related to large companies: What are the challenges large companies face when introducing AI? And if you were appointed head of AI at a Fortune 500 company, where would you start to bring value?
Jeff: If I were the head of AI at a Fortune 500, I would mandate that all top AI capabilities be available to everyone in the organization. I would release the constraints many companies put on employees—don't experiment, don't risk proprietary data getting out. I want more experiments and more initiatives.
Then I would create a reward system for the most innovative AI applications, running pilots, and creating a pipeline process to take successful pilots into production once they hit certain metrics. We call it "pilot hell" when our portfolio companies run successful pilots, hit the metrics, go back to their champions and ask, "What's next?" only to hear, "We hadn't thought about that—we need committees to figure out enterprise-wide deployment."
I would pre-wire the enterprise-wide playbook so that when a pilot hits its threshold, you can immediately deploy across the enterprise. So, not only more experiments but faster enterprise-wide deployment.
As for obstacles, I mentioned pilot hell—that gap between pilots and deployment. The other major challenge is training. If you have a 50,000-person company, how do you train everyone to be AI-native? The best big company CEOs model their behavior, like Marc Benioff at Salesforce. You need to make sure your entire organization knows this is a priority, not a toy or a fad, but a fundamental competitive advantage.
Alvaro: There's a question from what I imagine is a private equity investor: How should private equity be thinking about AI?
Jeff: Great question. One of our LPs runs a private equity firm and hired an AI expert to help work through their portfolio. Many PE firms have portfolio support teams, and this AI leader sits alongside the talent leader, marketing leader, and financing leader to look at internal tools and work with portfolio companies.
The really interesting opportunity I'm seeing is building AI-native service firms instead of trying to sell software to make services more efficient. We recently invested in an AI-native law firm that uses these tools to be more efficient in drafting contracts and conducting reviews. You see similar developments in accounting, finance, and consulting fields where individuals earning $200-300K are doing repeatable, automatable tasks and applying judgment that can be replicated with AI tools. This will be a fascinating area for both VCs and private equity.
Alvaro: One last question: At the elementary and high school level, how are we preparing new generations for this new world?
Jeff: Last year at Harvard Business School, our students were told, "Be careful, be cautious, don't violate rules." This year, they're being told "embrace, extend, experiment." There's been a massive cultural shift, and I'm seeing it in many schools, though it will take longer in some places.
A founder friend recently told me about his 14-year-old son who loaded all his study guides, textbooks, and lesson plans into notebook LM and asked it to develop sample tests to prepare for his midterm. After taking these practice tests, the AI identified knowledge gaps and created personalized audio tutorials for those areas. This dramatically accelerated his learning.
I think that's the future we'll see in schools—personalized tutors and tools for more effective knowledge transfer and coaching. Khan Academy has been brilliant in this area. We're heading toward a world where everyone has a private tutor who knows your history, understands your background, and helps with lifelong learning.
Alvaro: This is a great way to end our fireside chat. Jeff, this is a perfect example of an AI-native person bringing these tools into their home and parenthood. There's nothing more important than family and children, and leveraging these tools for better education shows how we can start thinking about AI in our everyday lives. Thank you so much, and thanks to the HBS Club of Mexico and the Rock Center for Entrepreneurship for co-hosting this event. I've learned a lot and hope it's been the same for everyone who joined us.
Jeff: Thanks, Alvaro, great to see everybody.