How Reforge Transformed From Education Company To Multi-Product Platform (Without Breaking Everything)

Brian Balfour on the organizational reality of shipping 5 AI products in 9 months, the 75% team loss nobody talks about, and why Discovery is the next bottleneck.

Every Product Manager I know is being asked the same question right now: Should we build AI tools or teach people how to use AI?

Reforge’s answer: Both. Simultaneously. With 25 people.

For seven years, Reforge was THE brand for product management education. Thousands of PMs learned how to build products through their cohort-based courses. Then, in 2024, they shipped five AI products in nine months while keeping the education platform running.

Build. Insights. Interviews. Launch. Artifacts.

Most companies struggle to ship one product without breaking everything else. Reforge shipped five and avoided the “Frankenstein workflows” that kill most multi-product strategies.

I wanted to understand how they actually pulled this off. Not the polished LinkedIn narrative, but the messy reality. The team conversations nobody shares publicly. The hard decisions about who stays and who goes. The moments of doubt at 2 AM.

So I asked Brian Balfour, Reforge’s CEO and former VP of Growth at HubSpot, to walk me through the organizational transformation that enabled this.

What he told me changes how I think about AI product strategy entirely.

Why 75% of Reforge's Team Couldn't Make the Transition

Let’s start with the number that made me pause the interview.

Elena: You went from a team optimized for creating courses to a team shipping software products. What percentage of your original team could make that transition?

Brian: Hard to estimate. I’d say about 25% of the team. To be clear, for the 75% it had nothing to do with their talent level. Different people are good at different things with different interests at different stages. They are all amazing people.

25%.

That’s the number most founders won’t admit publicly. Three out of four people on the original team couldn’t transition to the new reality. Not because they weren’t talented, but because the skills required were fundamentally different.

Think about what this means practically: If you’re a 20-person company transforming into a product company, you’re going to lose 15 people. Not because they’re bad. Because the job fundamentally changed.

This is why most education companies that try to build products fail. They try to transform the entire team instead of building a new team around the 25% who can make the leap.

Reforge made two acquisitions specifically to bring in founder-level talent. That’s not a nice-to-have. That’s the strategy.

💡 Key Insight: Company transformation isn’t just about strategy and technology. It’s about accepting that 75% of your team might not be able to make the leap. The question isn’t whether this will happen. It’s how you handle it when it does.

When the CEO Becomes an Individual Contributor

Most management books tell CEOs to “stay strategic” and “delegate everything.” Brian did the opposite.

Elena: When you’re shipping 5 products with 25 people, someone’s wearing multiple hats. What roles did YOU have to personally take on that you never expected?

Brian: I stopped managing and went full IC (individual contributor). I gave all people, operational, and other responsibilities to our amazing COO Tom Willerer and inserted myself wherever we needed the most activation energy.

Something I learned during my time at HubSpot is that when you are launching new products within an established company, the temptation is to skip steps. You think you can because you have more people, money, distribution, etc. You can’t. Those things only help you flow through the steps faster. Skipping leads to failure.

That means I dug in with some of the product teams. We did founder-led sales first before handing it off to others. Created all the sales and marketing materials myself initially. These are things you can’t skip.

But there’s something deeper Brian learned through this process. Teaching product management for seven years didn’t make building products easy:

“You can have more knowledge than anyone else, but that does not ensure success. Success comes down to combining knowledge with many other factors, like having the right people with you, getting timing right, choosing the right market, and plenty of luck along the way.”

This is the humility that comes from eating your own dog food. Reforge built their brand teaching PMs how to execute. Now they’re proving that even when you know all the frameworks, execution is still hard.

Read that again. The CEO of Reforge stopped managing people and started writing sales decks.

This goes against every piece of leadership advice about “working ON the business, not IN the business.” But Brian’s right. When you’re launching new products, there are steps you cannot skip, regardless of your resources.

Someone has to do the ugly, unglamorous work of figuring out how to sell the thing before you can scale. And if you’re the CEO, sometimes that someone is you.

💡 Key Insight: Resources don’t let you skip steps. They let you flow through steps faster. The CEO writing the first sales deck isn’t a failure of delegation. It’s insurance against failure.

The Emotional Cost of Fast Decisions

Transformation at this speed isn’t just organizationally expensive. It’s emotionally brutal.

Elena: In the Supra interview, you mentioned avoiding “slow-to-die experiments.” How do you actually kill experiments at Reforge?

Brian: The criteria is tailored to the initiative and experiment. It’s fully dependent on what you are trying to prove. But once you decide, the only way to do it is fast and swift. Yes, some feelings will be hurt. But people will get over it. We are here to take a shot at building something big. That is the only thing that matters.

This is the part of product management that nobody teaches in courses. How to kill things quickly when they’re not working. How to have the hard conversations. How to accept that being fast means being uncomfortable.

Most PMs try to soften the blow. They give experiments extra time. They let things die slowly to avoid the emotional cost of pulling the plug.

Brian’s approach: Rip the bandaid off. Accept the emotional fallout. Move on.

If you’re planning your 2026 AI roadmap, this ruthlessness about killing experiments might be the most important skill you develop.

Why Discovery is the Next Bottleneck

Shipping Faster Makes Bad Products Easier

Here’s where the conversation shifted from organizational tactics to strategic thesis.

I’ve been writing about AI evaluation systems and how teams struggle to ship reliable AI features. Brian’s arguing that evaluation isn’t even the biggest problem.

The problem is further upstream.

Elena: You’ve framed AI adoption as three pillars: Discovery, Delivery, and Adoption. But YOU’RE building tools for all three pillars simultaneously. How do you prevent creating the same bottleneck problem you’re solving for customers?

Brian: We are actually much more focused on Discovery. We believe this is the next most important bottleneck. Code and designs are liabilities that need to be maintained. If you ship a bunch of stuff that only a few percent of users use, you are doomed over the long term. AI is making it easier to accrue these liabilities.

Many folks will shift towards “we are shipping tons of stuff, but now we need to fix adoption.” To solve that they will do a bunch of wonky stuff like tool tips, pop-ups, and never-ending lifecycle emails.

But the real problem is typically further upstream. You are deciding to ship a ton of stuff customers don’t actually care about. Discovery is still an extremely fragmented process across qualitative information, quantitative information, internal context, ideation, internal feedback, external feedback, etc.

Let me unpack why this matters so much.

Why You Can Now Ship Bad Ideas 10x Faster Than Before

Right now, every AI narrative is about velocity. Ship faster. Iterate faster. Move faster.

Tools like Cursor, Claude Code, and v0 let you go from idea to working code in hours instead of weeks. That sounds amazing until you realize what it actually means:

You can now ship bad ideas 10x faster than before.

When shipping was slow and expensive, you were forced to be thoughtful about what you built. The friction was frustrating, but it created a natural filter.

AI removes that friction. Which means teams are accumulating “code and design liabilities” faster than ever.

Brian’s thesis: The competitive advantage isn’t shipping faster. It’s discovering better.

When everyone can ship quickly, the winners will be the teams who figure out what’s worth shipping in the first place.

This is why Reforge Build focuses on Discovery, not Delivery. They’re betting that prototyping and validating ideas before committing engineering time will matter more than coding speed.

💡 Key Insight: AI doesn’t make bad products disappear. It makes them easier to build. The bottleneck shifts from “can we build this?” to “should we build this?” Discovery matters more, not less.

Why Reforge Build Isn't Competing with Cursor or v0

I pushed Brian on this. Reforge Build is competing with Lovable, v0, Bolt, Cursor. All well-funded AI coding tools. Why would anyone choose Build?

Brian: There are two main points of differentiation for Reforge Build:

First, Reforge Build is specifically designed for product teams working on existing products. That is very different than the app builders who are optimizing for zero-to-one use cases. We are trying to make it as easy as possible to prototype from your existing designs and context (strategy, customers, etc).

Second, a prototype is just an artifact. There is tons of workflow around that artifact. That workflow is different based on what you are using the artifact for (internal discussion, customer validation, etc). We think these workflow pieces will be more and more important long term.

This distinction is crucial.

App builders (Lovable, v0, Bolt): Start from scratch. Build entire apps. Optimize for founders.

Reforge Build: Start from your existing product. Create variants. Optimize for product teams.

Most PMs aren’t building new apps. They’re deciding what to add to existing products. Build addresses that specific use case.

But the second point is even more interesting: “A prototype is just an artifact.”

Brian’s betting that the workflow around prototypes—sharing with stakeholders, getting feedback, documenting decisions, running user tests—matters more than the prototype itself.

This is the same pattern HubSpot used to dominate CRM. Salesforce sold you a database. HubSpot sold you a complete workflow for managing customer relationships.

If Brian’s right, Reforge won’t compete on “who has the best AI prototyping tech.” They’ll compete on “who has the best workflow for making product decisions.”

How to Ship 5 Products in 9 Months: Hire Founders and Let Them Cook

Okay, but HOW do you actually ship five products in nine months with 25 people?

Elena: What’s ONE process or principle that made this possible that other founders could steal tomorrow?

Brian: First, you need other founder-type talent in the company to do this. We made two small acquisitions in late 2024 to bring this talent in. Today, there are 7 former founders in the company including myself.

Second, let them cook. Set a direction and then get out of their damn way. You need to trust. This is something I see a lot of founders struggle with. We could not have done this without some amazing people like Chun Jiang, Ben Kramer, Jeff Dwyer, Adam Green, Dan Wolchonok and others.

That’s the entire strategy. Hire founder-level talent, give them direction, then get out of the way.

This sounds simple until you realize how few founders can actually do this. Most want to stay involved in every decision. They say they trust their team, but they’re in every Slack thread, every design review, every product spec.

Brian’s approach requires a level of delegation that most CEOs aren’t comfortable with.

But it’s also the only way to move this fast.

Think about it: If Brian tried to be involved in decision-making for Build, Insights, Interviews, Launch, AND Artifacts, nothing would ship. He’d become the bottleneck.

Instead, he hired 7 former founders—people who’ve run their own companies and made their own product bets—and trusted them to execute.

💡 Key Insight: You cannot ship multiple products simultaneously unless you have multiple founder-level decision-makers who can operate independently. Acquisitions aren’t about hiring talent. They’re about importing decision-making capacity.

The Doubts Every Founder Has (But Won’t Admit)

I saved the hardest question for last.

Elena: You’ve built Reforge into THE brand for PM education. Now you’re competing with AI coding tools, prototyping platforms, and feature management systems. Does part of you worry that by expanding beyond education, you’re risking the thing that made Reforge special in the first place?

Brian: Yes, of course. A venture-backed founder’s job is to take bets that are typically high risk, high reward. There are a lot of people who aren’t up for that, which is completely fine. But if that’s the case, then get off the venture startup track.

Founders are incredibly high conviction on the surface. You need to be in order to rally the money, people, and energy to have a chance. But deep down, every founder I know has doubts and concerns. Dealing with those doubts is half the battle.

This might be the most honest thing Brian said in the entire interview.

Every founder projects confidence publicly. But privately, they all have doubts.

The question isn’t whether you have doubts. It’s how you deal with them while still moving forward.

Brian’s not pretending the risk doesn’t exist. He’s not spinning it into “actually, this makes us stronger.” He’s saying: Yes, this is risky. Yes, I have doubts. And we’re doing it anyway because that’s the job.

Seven Years of Customer Feedback Finally Made Sense When AI Arrived

We’ve covered what Reforge did and how they executed. But why did they do it in the first place?

Elena: There’s a specific moment when Reforge stopped being purely an education company. Can you walk me through the actual conversation or realization that triggered “we need to build tools, not just teach about them”?

Brian: It’s never one conversation or one moment. It’s a series of conversations and data points that build up over time in three places:

1. Customer problems and opportunities

2. Business problems and opportunities

3. Market signals and trends

The most important thing is that it was rooted in something we had been hearing from customers for years: “That course was amazing…but can you help me do/implement X.” We had heard that over and over for 7 years, but it never made sense for us to focus on and solve until AI started to inflect.

AI triggered a couple of things:

First, a new market need. Every company is going through both a technology and a knowledge/behavior shift.

Second, new product possibilities. You can now “bake” knowledge directly into the tools themselves.

This is the pattern I see most founders miss. They wait for the singular “aha moment” that never comes.

Brian and his team had been collecting signal for seven years. Thousands of students saying “this is great, but can you help me actually implement it?”

The pivot wasn’t impulsive. It was inevitable once the technology caught up to the need.

This is what real product strategy looks like. Not reacting to market trends. Not chasing the hot new thing. Listening to customer problems for years, then acting when the technology finally enables a solution.

💡 Key Insight: The best product decisions aren’t made in a single meeting. They’re made through years of listening to customer feedback and waiting for the right technology moment to act on it.

Are Reforge's Students Actually Buying Their Tools?

On paper, Reforge’s strategy is obvious. Teach PMs how to build products, then sell them the tools to do it.

But does the flywheel actually work? Or are they separate audiences?

Brian: Education is both driving tool adoption and they are separate audiences. AI is both a tool change and a knowledge change, but those two things don’t necessarily happen at the same time.

This strategy has been around for 15 years. My alma mater, HubSpot, was the first to truly realize this. If your only customer touch point is when someone is “in-market” for a solution, you will lose. You have to have touch points outside of that.

Most companies are terrible at this because all they want to talk about is their product. But you need to constantly find ways to talk about and address your customer problems and needs even if they aren’t in market for a new solution. If you do that, then when they are in-market, you will be top of mind and in the consideration list.

This is the content marketing playbook that HubSpot pioneered and that Reforge is now executing at a different level.

The insight: Education and tools serve different needs at different times.

Someone taking a Reforge course might not be ready to buy Build. But six months later, when they’re planning their roadmap and need to prototype AI features, Reforge is the first name they think of.

This only works if you’re not constantly selling. You have to actually provide value through education, even to people who will never buy your tools.

Most companies can’t execute this because they’re too impatient. They want every piece of content to drive conversions.

Reforge plays a longer game.

What Actually Keeps Brian Up at Night: The F1 Speed of AI Markets

Elena: What’s the hardest decision you’re facing RIGHT NOW with Reforge that keeps you up at night?

Brian: See above on speed of AI market.

The brevity of this answer tells you everything.

Earlier in the interview, Brian talked about how “the hardest thing in practice has been the acceleration of everything in this AI environment. You need to hire faster, build faster, make faster decisions, market faster, etc. It’s like driving an F1 race car. The slightest bump or tap on the steering wheel can send you into a wall.“

The speed of the AI market isn’t just a challenge. It’s THE challenge. Everything else is downstream.

When I asked what keeps him up at night, he didn’t give me a specific decision. He gave me the underlying condition that makes every decision harder: Everything is moving faster than anyone is prepared for.

The Unsolved Retention Problem

In his Behind the Craft interview, Brian talked about how AI products struggle with retention. But Reforge is building AI products themselves.

So I asked him directly: How are you solving your own problem?

Brian: We haven’t fully solved it yet. Retention is a lifelong endeavor. Right now in most categories the market is still in an experimental mindset. Most individuals and companies haven’t settled in to a single tool. They are constantly trying new things. But that phase will not last forever. So we are building towards an audience and use case that we believe will have very high retention for when that dust starts to settle.

I appreciate the honesty here.

Most founders would spin this into “we’ve cracked retention with our innovative approach.” Brian admits they haven’t solved it yet. They’re building toward a thesis about what will drive retention, but the market is still too experimental to know if they’re right.

This is what real product strategy looks like. Not certainty, but informed bets based on market signals.

Reforge is betting that:

1. Product teams working on existing products will have higher retention than people building new apps from scratch

2. Workflow features (the stuff around the prototype) will create stickiness beyond the core prototyping feature

They might be right. They might be wrong. But they’re building toward a clear thesis instead of just hoping retention improves.

What This Means For You

If you’re a PM thinking about AI adoption, multi-product strategy, or organizational transformation, here’s what Brian’s journey reveals:

On Transformation:

• Expect only 25% of your team to successfully transition when you fundamentally change your company’s direction

• You cannot skip steps, even with resources. You can only flow through them faster

• Transformation requires founder-level talent who can operate independently

• Sometimes the CEO needs to become an IC again to avoid skipping critical steps

On AI Products:

• The defining challenge of the AI market is its speed; everything, including mistakes, moves faster.

• Retention in AI products is unsolved; build toward a thesis but accept uncertainty

• Code and designs are liabilities; AI makes it easier to accumulate bad liabilities faster

• When everyone can ship fast, competitive advantage shifts to discovering what’s worth shipping

On Strategy:

• Listen to customer problems for years before acting. The best pivots aren’t impulsive.

• Discovery is the next bottleneck as AI makes Delivery faster

• Throwing away work isn’t waste; it’s the path to great products

• Workflow around artifacts matters more than the artifacts themselves

On Leadership:

• Trust founder-level talent and get out of their way (”let them cook”)

• Kill experiments fast and accept the emotional cost

• Every founder has doubts; dealing with them is half the battle

• When transformation requires losing 75% of your team, own it and move forward

Why Reforge Won't Build a Coding Tool (Even Though Everyone Else Is)

One final question: If Build focuses on Discovery, are they tempted to build the Delivery layer too?

Brian: No. There are so many good tools there (Cursor, Claude Code, Devin, Codex, etc) that we don’t feel a need to play there. I don’t agree that discovery without delivery is just expensive experimentation. I think that can lead to the mindset that work/ideas you throw away in discovery is a “waste.”

This leads teams to not explore a wide enough territory to find the best solution which then inherently leads to a worse product. You have to accept that throwing away work is part of the process to getting to a great product. The lower the friction is to the discovery process, the more you can explore, the easier it is to throw things away, and get to the best possible outcome.

This reframe matters.

Most PMs view throwing away work as failure. Brian views it as a necessary part of finding the best solution.

The lower the friction to explore, the more you can explore, the easier it is to throw things away. That’s why Discovery matters more as AI makes Delivery faster.

Five Years From Now

Elena: Five years from now, is Reforge primarily an education company that builds tools, or a tools company that happens to teach?

Brian: False choice. The best software companies are those that enact large changes and transformations for their customers. That’s both a software and an education problem.

This is the answer I expected, but it’s still the right one.

The companies that win in AI won’t be pure tool companies or pure education companies. They’ll be the ones that understand transformation requires both.

You can’t just give someone an AI prototyping tool and expect them to use it well. You need to teach them how to think about Discovery, how to run effective user tests, how to make product decisions with incomplete information.

Reforge isn’t building tools OR education. They’re building the complete transformation system.

Final Thoughts

Reforge’s transformation isn’t finished. They’re building toward a vision of AI adoption that may or may not be right.

But they’re doing it with:

• Clarity about the trade-offs (losing 75% of the team)

• Honesty about the challenges (retention is unsolved, speed is brutal)

• Willingness to take the bet despite the doubts

Not certainty. Not guarantees. Just informed bets, ruthless execution, and the courage to keep moving forward when you’re not sure if you’re right.

That’s what real product leadership looks like now.

If you’re planning AI features for your roadmap, you can prototype and validate them before committing engineering time with Reforge Build. It’s designed specifically for product teams working on existing products, not another app builder for starting from scratch.

The difference between shipping AI that works and AI that fails often comes down to whether you validated the idea before you built it.

What transformation are you facing in your product organization? What’s your biggest challenge with AI adoption? Drop your thoughts in the comments.