Finding Product-Market Fit in the Age of AI, with Venture Capitalist Jeff Bussgang

Introduction

Elliott: Welcome to Minimum Viable Podcast, where we go beyond the surface to explore the depths of product creation. I'm your host, Elliott Poppel, founder of General Collaboration, a cross-inbox tool that keeps you focused on what needs your attention most.

^{The Entrepreneurial Journey}

Elliott: Jeff, I'm thrilled to have you here today. Our paths have crossed over the years—you invested in my previous venture, and now I'm working with your partner Jesse on General Collaboration. Your background is fascinating—from founding YouPromise to your two decades at Flybridge and twelve years teaching at Harvard Business School.

Jeff: It's great to join you, Elliott. I'm excited about this podcast as a thoughtful platform for product managers and innovation enthusiasts.

Elliott: Beyond investing and teaching, you've authored multiple books. What drove you to write?

Jeff: Writing wasn't in my career plan, honestly. It became my outlet for persistent ideas that demanded expression. My first book, "Mastering the VC Game," emerged from seeing founders struggle with the fundraising labyrinth. "Entering Startup Land" followed as a guide for professionals navigating startup careers. Now I'm completing "The Experimentation Machine: Finding Product Market Fit in the Age of AI," born from watching these powerful new tools reshape the entrepreneurial landscape.

_{AI's Impact on Product Development}

Elliott: At General Collaboration, we're integrating AI both behind the scenes and in our user experience. I'm curious—what sparked your exploration of AI and product-market fit?

Jeff: It was a natural convergence of my worlds. At Flybridge, we've invested in AI companies for over a decade. Meanwhile, my HBS course focuses entirely on the zero-to-one journey of finding product-market fit. I began noticing how founders using AI tools were accelerating through critical early stages, making discoveries faster and pivoting more efficiently. The more I incorporated these examples into my teaching, the more compelling the connection became.

Elliott: Many technologies come with hype cycles. What convinced you that AI deserves special attention compared to blockchain or other innovations?

Jeff: Look at the word itself—artificial intelligence. The intelligence component is transformative for decision-making, which lies at the heart of entrepreneurship and product management. Every day, founders make countless decisions about features, hiring, and market approaches—often with limited data. AI doesn't just automate; it enhances the quality of those decisions through pattern recognition and insight generation.

I've come to believe that AI won't replace founders, but founders who effectively leverage AI will eventually replace those who don't. The same applies to product managers. It's not about surrendering to technology but amplifying your capabilities with it.

^{Practical Implementation}

Elliott: Is there a universal playbook for implementing AI, or should each founder chart their course?

Jeff: Think of it as applying timeless principles with timely tools. The fundamental entrepreneurial process—identifying pain points, crafting value propositions, running experiments—remains unchanged. What AI offers is unprecedented acceleration and insight throughout that process.

Start by asking: In your most critical current challenge, how might an AI collaborator help you move faster or see clearly? Perhaps it's synthesizing customer interviews, spotting patterns in user behavior, or generating creative alternatives to test. The entry point should align with your immediate priorities.

Elliott: That perspective resonates with me. As a third-time founder, I've learned to distinguish between productive tools and distractions. I literally keep a note on my desk saying "don't fill up on bread"—reminding me to save room for what truly matters rather than getting lost in peripheral activities, however enticing they might be_{Getting Started with AI}

Elliott: I worry sometimes that I look at AI as one of those distractions. How do I start? Where do you recommend that I start?

Jeff: It's a great question.

Elliott: Which tool should I start with? That's part of the problem. My buddies are saying I need to pay for Perplexity. Should I carve out a day, an hour?

Jeff: I would start with ChatGPT personally, and play with the idea of talking to it like you would a business therapist. Tell it your problems, tell it what you're struggling with, then give it information and see if it can help you. Start small with a particular thing.

For example, take customer discovery- everyone's always gathering customer insights. You might have interviews and interview notes. Ask it to synthesize your interview notes and derive insights, then ask it to recommend three features that might address the issues that come up. You start having this back-and-forth dialogue, treating it almost like a human, a co-pilot, or co-founder.

I've been playing around with evaluating investment theses. As an investor, my team and I are always developing various investment theses. In my class, I assigned students to create investment theses they found exciting. I trained the model on the best examples, then applied it to student submissions, asking for three strengths and three areas for improvement about each thesis. It provided very cogent feedback.

One of the first assignments I give my students is to generate startup ideas using ChatGPT in sectors they find interesting. If fraud in Medicare Advantage interests you, ask ChatGPT for startup ideas in that area. You can then ask it to play the role of a user persona - "Tell me what a claims processor in Medicare Advantage worries about, their workday, workflow, tools, priorities, and how they're compensated." You begin this inquiry process and suddenly have a mock user persona you can question, like having a customer sitting next to you.

_{Practical Applications for Product Teams}

Elliott: That's super interesting. For General Collaboration, we're focused on product people - product managers, designers, or developers at companies with 5 to 50 employees.

Jeff: If I were you, I'd immediately create a custom GPT using OpenAI's functionality and ask it to act like a product manager at an organization with 50 employees using certain tools with specific objectives. Then begin dialogue with it as if you had a full-time, 24/7 voice of the customer sitting next to you. It could be incredibly insightful for you and your team to bounce ideas off, like creating a synthetic customer advisory council.

Elliott: That's wild! We're a very user research-heavy company - I'm probably talking to 4-5 users daily. I've shied away from feeding all that data into AI. You just highlighted what a missed opportunity that is - feeding it all into creating a singular person I can talk to is pretty cool.

_{AI Applications for Early-Stage Companies}

Elliott: What else are you recommending people do at the early stages of finding product-market fit?

Jeff: One of my favorites is using mock user personas and synthesizing user feedback to drive insights and recommend features. Another is the process of qualifying target customers. Train the AI on your ideal customer profile (ICP), then have it develop five questions to help score whether a lead is top-qualified, medium-qualified, or low-qualified. This helps your sales reps spend their time on the right users.

It's also incredibly effective at content generation - email templates, blog posts, and SEO content. One of my portfolio companies uses it to generate personalized videos for target customers. If your sales team wants to reach out to 10-20 targets, you can use generative tools to create personalized videos based on their website and social media, generating a 15-20-second video of you reaching out to a particular prospect.

_{Team Structure for AI Implementation}

Elliott: Is my team properly staffed to address this? We have limited engineering bandwidth. Are you seeing teams change how they hire at early stages? Can I do this as a relatively non-technical person?

Jeff: Great question. Ideally, you'd begin with no-code tools - ChatGPT has fantastic no-code capabilities. Your non-technical team could start experimenting with these approaches. As you advance, someone on your engineering team or perhaps a technical non-programmer could build custom chatbots or workflows.

It's important that your engineering team experiment with these tools and APIs to build pipelines. For example, one of our portfolio companies, an enterprise software company that responds to many RFPs, had an engineer automate this process for their sales team. Using OpenAI's API, he created an automatic RFP response generator trained on previous proposals. When fed a new request, it automatically generates answers that get you 80-90% of the way there, which you then edit.

I just wrote a blog post about this after reading Ethan Mollick's new book "Co-Intelligence." I keep coming back to that word: intelligence.

_{AI Implementation Framework}

Jeff: Ethan Mollick's new book "Co-Intelligence" offers a useful framework for AI tasks. As a leader, consider what types of tasks you want AI to handle:

1. "Just Me" tasks - Things only you can do (like making hiring decisions)

2. Delegated tasks - Tasks where AI can do 80-90% and you provide final oversight (like RFP responses)

3. Automated tasks - Completely automated processes (like personalized outbound emails)

In your daily work, consider: "If I had a wise, intelligent AI next to me, what would I delegate, what would I automate, and what would only I do?"

_{Common Pitfalls to Avoid}

Elliott: Where are you seeing companies waste time with AI?

Jeff: I've seen companies waste time with analysis paralysis - trying to optimize everything perfectly, debating which LLM to use, or considering complex programming with open-source models like Llama. They spend too much time analyzing choices rather than diving in and experimenting.

It's similar to what we tell founders about starting companies: put an MVP out there, see what sticks, and observe user reactions. For internal AI prototypes and efficiency tools, just start playing around - don't let paralysis by analysis cause you to freeze.

_{Untapped Opportunities}

Elliott: What else should I be doing that's unique or off the beaten path? What are the most interesting implementations you've seen?

Jeff: Customer service is an area ripe for automation and AI co-pilots. Even if not directly applicable to you, many companies build up customer success functions that could benefit from customized chatbots for FAQs or tools that automatically generate email responses.

I like to think of it as creating a "department of one" - can one salesperson with AI co-pilots have the impact of an entire department? Can one customer service rep or product manager achieve the same? This lets founders focus on the "just me" tasks that truly require their attention.

Another example is that many founders gather analytical information from their websites, financial systems like Stripe, or tools like Mixpanel. You can automatically generate board decks, financial summaries, and investor updates from this raw data.

_{Tool Recommendations}

Elliott: With limited runway and cash to burn, are there free tools to try? Are there any must-have applications?

Jeff: ChatGPT Pro is a must-have. I'm not an investor in OpenAI, but having access to the pro version is powerful. For product managers, Chat PRD helps generate requirements documents through dialogue, getting you 80% of the way there - I believe it has a free version.

Many tools offer free trials with premium versions. Custom GPT is good for creating chatbots for FAQs. HeyGen is nice for video generation, like those personalized videos for leads. Many of these tools are affordable at $5-20 per month to get started.

_{The Future of AI Interactions}

Elliott: At General Collaboration, we're considering AI features for writing and reading comments. I'm concerned about scenarios where my bot talks to your bot, like my AI creating a pitch deck hitting every buzzword from your website, and your AI reading it. That kind of scares me.

Jeff: That doesn't scare me at all - it excites me! It would cut through the noise and make everyone more efficient. Imagine as a founder analyzing every investment that every VC firm has made - stage, size, sector - and having it identify the 10 firms and partners who are the best match for you. That would be amazing and save tremendous hassle with filtering and qualifying.

It's interesting - you're a tech founder building a new software product, trying to get people to experiment with it and overcome fear of the unknown. All founders doing that should, in turn, embrace these new tools themselves.

Elliott: I know I need to overcome that fear and experiment more. We're using AI in interesting ways for product development, but it's more of a human question. Especially when building tools for collaboration and communication, it concerns me if we're creating content no one reads, like PRDs that AI generates and AI reads. Maybe the process evolves beyond traditional documentation.

Jeff: It's a fascinating idea to consider the entire pipeline. I feed priorities into the PRD generator, which creates the PRD, then the engineering team feeds that into their generator...

_{The AI-Enhanced Development Pipeline}

Jeff: The engineering team could feed that PRD into their generator, which produces actual code and features reflecting the PRD. Much of this will get you 70-80-90% of the way there, with humans coming in to edit, review, and refine.

Looking at current AI outputs in music, video, and prose, they're good, sometimes passing the Turing test, but usually it's obvious they're templated and automated. Still, they get you started and significantly down the path.

Here's a key point, Elliott: We're playing with these tools as of April 2024, but if you establish these capabilities and processes internally now, you'll be positioned to take advantage as tools improve. When ChatGPT-5 releases in the coming months, you'll have already laid the groundwork. Founders, product managers, and executives must start experimenting with these tools even if results are imperfect, because developing that muscle will be incredibly valuable as underlying platforms improve.

_{AI's Impact on Investment Strategy}

Elliott: Shifting gears, let's talk about your investor perspective. I'll need to raise our seed round for General Collaboration sometime soon. The world has changed since I initially pitched GC - we didn't have to be an AI company then, but now it seems every company is. How has your investing approach changed in the last 12-18 months because of AI?

Jeff: We believe the best companies will be those leveraging AI to be more effective and efficient. Even if you're not providing an AI offering, we'll be looking at whether you're using AI tools effectively.

Imagine two companies we discuss in a partner’s meeting. One describes business processes, customer discovery, and product development in ways that sound very "2010." The other showcases incredible mechanisms they're using to be more efficient in product development, go-to-market, and business model quality, and they're moving much faster with fewer people. We’ll be more excited about the second company, regardless of sector or founder background.

The burden is on founders to not only be in hot spaces - I think of founders as heat-seeking missiles finding customer pain or market momentum - but also to be savvy about leveraging the most modern, valuable tools in their startup journey.

Elliott: That's reassuring. There's a belief among founders that everyone needs to become an AI company, but you're saying we need to be savvy adopters of AI, run capital-efficient businesses, and maximize runway by using these tools, not necessarily be an AI product?

Jeff: That's right. There will be non-AI products in the world, but there won't be successful companies that avoid using AI to deliver their products. The best companies will effectively leverage AI, just as 20 years ago, you could predict that the best companies would be those most effectively leveraging the internet for distribution, communication, and internal efficiency. It's the same in this AI wave.

_{Disruption and Competition}

Elliott: This sounds like the classic innovator's dilemma.

Jeff: You're citing Clay Christensen, my late faculty colleague at Harvard Business School, who wrote about how competition from below, through technology shifts like AI, internet, or cloud, can deliver products at lower costs. Initially, they might be ignored because of lower quality, but over time, they leverage innovations to move up the quality chain.

What's interesting about this moment is that the big companies understand the shift. Microsoft isn't asleep at the wheel as it arguably was during the Internet and mobile eras. Google and Amazon are fully engaged. These companies are investing tremendously to provide AI tools.

I like to think of this as an "AI dividend" for founders, similar to the peace dividend after the Cold War. All founders are benefiting from this infrastructure, tools, and massive investments in capabilities. The best founders will know how to leverage this dividend.

Elliott: Does that impact the classic investor question, "What if Google does this?" Previously, I might have dismissed that concern, but now these companies are doing those things.

Jeff: It is changing how we invest. We always ask about competitive moats and sustainable competitive advantages, and the bar has gotten higher. You have to ask: What's proprietary? Is the dataset proprietary? Are workflows proprietary? Are relationships proprietary?

With Big Tech so sophisticated in this new wave, what will the next release provide? That's why at Flybridge, we're more excited about the application layer and tooling around the data stack. We were successful with our investment in MongoDB during the cloud era, and with companies like Zest AI, which leads in AI for bank underwriting. Having targeted, focused application areas and customers tends to be what we find exciting as investors.

Elliott: The landscape for acquisition has changed, too. Previously, I might have said we'd get bought by Google or Apple someday, but now that Instagram-style acquisition seems less likely...

_{Key Investment Perspectives in Today's Market}

Elliott: Has the current political economic climate changed your approach to investing? Are you looking at companies that can stand on their own? Do you still see acquisition as a viable outcome for many startups, or is that kind of off the table these days?

Jeff Bussgang: It's a very nuanced question. First, you have to step back and ask how investors think about their investments. We tend to be believers in the power law, which means our results will be defined by our massive winners, not by losers or companies that just do okay. You want to have a few massive winners in a portfolio, so every investment we make, we ask ourselves: "Could this be an outlier? Could this be a massive winner?"

To be a massive winner, typically you have to be a public company. So we really push ourselves to think, "If everything went well—if things work out the way this founder hopes—could this be a public company?" That's sort of the basis. Then yes, if things don't work out well, maybe the company gets acquired, or maybe it gets acquired along the way.

You're pointing to a delicate situation politically. The current Administration is pretty careful about antitrust, and there's a lot of friction right now in large acquisitions. But for us, those tend to be such rare events that would trigger an antitrust concern, so we tend not to focus too much on it.

We're more focused on whether this can be an independent, standalone public company. At that point, once the company completes its public offering—probably valued in the $2 to $10 billion range—early investors distribute the shares, and whatever happens next is in the hands of the public markets.

Elliott: So what you're saying is you only ever cared about home runs and grand slams. The singles, doubles, and triples were nice, but they were never part of the model to begin with. And yes, it might be harder to get acquired, but that wasn't something you were planning for anyway because you wanted those outsized outcomes.

Jeff: It's not something we're planning for, and I don't know if it's harder. On one hand, there's this headwind about antitrust, but on the other hand, there's a tailwind: acquirers are more numerous, way bigger, have larger market caps, and can spend more money more easily.

Today, for Google, a billion-dollar acquisition is not as big a deal as it was 10 years ago. For Meta, your comment about Instagram—that acquisition at a billion dollars 10-12 years ago wouldn't be as big a deal for them today with their market cap approaching $1 trillion, whereas at that time, their market cap was around $100 billion or $50 billion.

So we can see M&A activity in our portfolio at even richer levels, and founders are arguably operating in a bigger economy and a bigger value creation ecosystem today than before.

_{Current Investment Focus}

Elliott: We talked about the political and economic climate and AI. What else are you looking for from the investing side in this particular moment?

Jeff: One thing that's clear about this moment is that it's incredibly dynamic. We're looking for founders who can operate in a dynamic environment, which we refer to in shorthand as their "clock rate."

Clock rate has two meanings: one is the raw horsepower and intelligence of an individual—that's my earlier comment about AI helping founders improve their performance. But secondly, it's the ability to act quickly. When you have a more dynamic environment, the founders that are acting quickly and shifting their organizations on a dime—when a new release comes out, they're jumping on it—those are the attributes we're looking for.

We're putting an even higher value on founders who are nimble and can operate with a very high clock speed.

Elliott: Anything else that you're looking for in particular these days?

Jeff: One question we often get is how technical founders need to be in this environment. You mentioned it yourself—you're not a technical founder, you're a product founder. On one hand, these new tools are allowing business founders and product founders to build prototypes, mockups, and actual early MVPs without an engineer. On the other hand, if you believe these are a new set of tools that will improve the odds of success, you want people who really know how to use them.

We are looking for founders that, if they're not technical, don't need to be PhDs in machine learning, but they need to be facile in leveraging the tools. As I said before, founders who use AI effectively are going to replace founders who don't use AI effectively. We're looking for founders who know how to use the tools and are very rapid and nimble in running those experiments and leveraging the latest toolset.

Elliott: Jeff, this has been such an insightful and actionable conversation. I feel like I need to rush off this call and download a bunch of stuff to play around with. You've given me some really specific ways that I can start applying AI where before I felt like I was just swimming at it—there were so many things that I didn't know where to start. Thank you so much for coming on the podcast and just for catching up today. This was great.

Jeff: My pleasure. Great to see you. I'm excited to see where you take it. Thanks a lot and have a great rest of your day.

Elliott: Thanks, take care.