Brian Balfour
Founder and CEO of Reforge, former VP of Growth at HubSpot
Where to hear and read about Brian Balfour2 episodes and tests
Lenny's PodcastWhy ChatGPT will be the next big growth channel (and how to capitalize on it) | Brian Balfour
Lenny's PodcastBrian Balfour: 10 lessons on career, growth, and lifeBrian Balfour is founder and CEO of Reforge and a former VP of Growth at HubSpot, who has watched Facebook, Google, iOS and LinkedIn each open and then close their platforms. The one thing most worth learning from him is to read distribution shifts as a predictable cycle, join early with one focused bet, and plan your exit from day one.
it's about being specific about who you aren't solving for
Brian Balfour: 10 lessons on career, growth, and life
Brian Balfour: the story and the lessons
1Brian Balfour's path from HubSpot growth to Reforge
Balfour ran growth at HubSpot and co-founded three other startups. He was VP of Growth at HubSpot and co-founded three other startups before becoming founder and CEO of Reforge.
His lessons on growth, hiring and planning come from that mix of building growth at HubSpot and founding companies.
He lived a platform's glory days and horror days early on. In one of his first companies he built on Facebook's platform and saw both the boom and the crash when it closed.
He has also watched Google Ads, SEO and the Apple App Store rise and fall, which is why he treats platform shifts as a pattern.
Reforge began as courses and added software once AI made it possible. Learners kept asking for tools to implement what was taught. Reforge said no for a long time, until AI made it possible to encode that knowledge into software.
Reforge Insights now gathers feedback from every source, finds the gaps and auto-generates research to fill them. For ten years companies also came to Reforge to drive transformations, and AI adoption is the latest one.
2Balfour on distribution and the four-step platform cycle
Balfour says a great product is necessary but not sufficient. Winners are the teams that get distribution before incumbents can copy them. He has spent years pushing back on advice that boils down to product quality.
He calls this race for 'escape velocity' a good summary of what a startup does, an idea Alex Rampell wrote up around 2015.
The race for distribution is harder now than before. Incumbents copy faster, search clicks and social reach are shrinking, and AI code generation multiplies competitors in every category.
New platforms are one of the main ways out. Cursor overtook GitHub Copilot in nine months or less.
Every platform runs the same four steps. Step zero is a fierce market with five to seven players. Step one is finding the moat, step two is opening a third-party platform, and step three is closing it.
Closing means shutting the platform down, absorbing top use cases as first-party apps, or suppressing organic reach to push paid. Facebook, Google, iOS, LinkedIn and Udemy all followed it.
He sees the cycle as competition, not malice. Platforms must keep growing and must avoid enabling their own disruption, so they close. Sometimes closing is better for users, as Facebook's channels did get spammy.
His view: "They're playing you. So you need to play them."
The cycles are getting shorter, so there is less time to play. Facebook's cycle took about five years and Google's took many more. Facebook opened in 2007 at about a quarter of MySpace's size, then took a cut, suppressed organic reach and absorbed uses like events and photos.
Udemy started with about 80% of revenue going to creators and later cut that share sharply.
Balfour predicts winners by retention, not by reach. Facebook was smaller than MySpace and Friendster in 2007 but retained better. Google and Facebook won on retention and engagement, not on the biggest distribution at the time.
"it was never the person who had the biggest distribution at the moment of time. It was the one that had the best retention and engagement."
3Why Balfour expects ChatGPT to be the next big channel
Balfour expects ChatGPT to open a third-party platform within about six months. He thinks AI has had a technology shift but not yet a distribution shift. A couple of years in, he sees all the ingredients for a new platform in place.
Casey Winters made the point that the two shifts rarely arrive together and the distribution shift comes a little later.
He argues the AI moat is context and memory, not the model. Leading models give about the same result, so the difference is which one holds more of your context. The more you use it, the more memory it stores, which feeds personalized outputs and another flywheel.
ChatGPT shipped memory early and keeps investing in data connectors, which he calls really context connectors.
Retention data backs his ChatGPT pick over Gemini's reach. Menlo Ventures data shows ChatGPT's retention curves level off higher than rivals' and form a rare smile curve. It has roughly 10x Claude's monthly users.
He reads the retention shift as the effect of memory. He wrote a post arguing against his own pick and says he may be wrong on ChatGPT but is confident the cycle will repeat.
He expects preferred partners first, then an open platform. The sequence he predicts is a launch with 10 to 20 brand-name partners that lend credibility, then a wider opening where the value exchange is defined.
Signals he cites include agent platform job postings, agent mode, shopping built into the interface and deeper attribution in results. He would bet a large portion of his net worth that OpenAI is already meeting potential partners.
He expects specialized agents because horizontal ones are hard to adopt. Users struggle with tools that do everything, and specific use cases need their own UI and data. That is why third-party agents should follow ChatGPT's general agent mode.
He also expects outcome-based agent pricing to erode as rivals undercut it, unless it is paired with a moat like proprietary data.
4Balfour's rules for betting on a new platform
Balfour says never opt out of the platform game. Skipping it means competitors join and customer expectations shift to the new experiences. He calls it a prisoner's dilemma, and being early beats being late.
HubSpot joined ChatGPT's connectors even though exposing data looks irrational in isolation. ChatGPT now sends Lenny's newsletter more traffic than Twitter, so blocking it would only hand that share to someone else.
Startups should pick one platform and go all in. Late-stage companies can spread bets and wait for a winner. Startups lack the resources, and the failures tend to be those playing several platform games at once.
On mobile, betting only on Android probably lost, while iOS alone could still win. A winning strategy had to include the platform that won.
He picks platforms by retention, user quality, value exchange and scale. In that order. Scale decides only when the gap is huge, for example 200x.
Android has 70-some percent of devices but only about 30% of dollars, and iOS is the exact flip.
He says to plan your exit the day you enter. Build defensibility by owning a workflow, accumulating specialized data and context the big platform lacks, or building micro network effects.
You may not know the exit at entry, but figure it out along the way rather than waiting for step three.
He tells teams to be ready to turn strategy on a dime. New platforms must be exploited extremely quickly. The hard part is often leaders' reluctance to cause whiplash across projects already in flight.
Until the value exchange is known, he suggests working out how you would evaluate each player's platform and building relationships with the owners.
5How Balfour thinks about growth, use cases and focus
Balfour treats growth as one system, so causes sit elsewhere. Retention problems often come from acquiring the wrong users, and monetization problems from weak engagement. Teams skip asking which part of the system causes the issue.
Early HubSpot had weak renewals because sales sold to the wrong persona and use case. Removing commission on those deals changed behavior at once.
He defends good friction even when it cuts early conversion. The right friction improves downstream engagement and metrics, but short-term wins get applauded, so friction gets chipped away.
Reforge refined its application questions and turns away eager applicants likely to have a bad time and leave bad reviews.
He defines products by use case, then asks who has it. A use case is the problem, the value prop, the alternative and why they choose you. The natural frequency of the problem sets retention and activation metrics, and frequency of adoption sets acquisition.
Reforge calls this the use case map, developed with Casey Winters and Shaun Clowes. Students usually map about eight use cases, which is far too many.
He builds a relationship with buyers before they are in market. Most SaaS buyers are in market only once every few years. HubSpot's content made many people think of it as a blog, so it was the first place they went when ready.
He names who he is not solving for. Early HubSpot found four use cases hidden behind one marketing customer, chose only the mid-market all-in-one 'Marketing Mary', and aligned sales comp, success and marketing to turn the other three away.
The idea came home from his summers as a wedding planner in LA, where the most miserable couples tried to please everyone. Reforge's values doc spends half its length on what each value does not mean.
He studies what everyone does in order to do the opposite. When online courses were cheap, self-serve and open, Reforge launched high-priced, small-group, live and selective, and cohort-based courses became a thing afterward.
He applies it to ad creative too: review every competing advertiser's creative, then build something completely different. When copiers arrive, flip the pendulum again.
He wrote long, infrequent content while HubSpot's norm was high-volume. The playbook then was short articles, around ten a day. Most of his early email subscribers came from his very long posts.
He notes the HubSpot approach worked for HubSpot because of its early domain authority.
6Balfour on getting real AI adoption inside companies
Balfour says hard constraints beat memos and incentives. One company capped each function at a fraction of peer team size for its revenue stage, so nobody could hire above it. Another rule is no new headcount until you prove AI can't do the work.
A smaller version is an executive who won't review a PRD without three prototypes. Clear messaging, owners, career ladders and review questions help, but he says constraints are the most impactful lever.
He sorts teams into catalysts, converts and anchors. Catalysts experiment on their own time, converts need structure, permission, budget and rewards, and anchors drag their feet and create silent friction.
Fewer than 10% of companies give anchors a date to change or leave, yet they are furthest along. His reasoning: "Cultures thrive on density."
He tells executives to audit adoption on the ground floor. About 90% of the time, a prototyping tool user said only they and one other person use it. Shopify measures real usage to stay close.
At one tech company a principal PM's prototype stalled for a month until the CEO heard about it at a happy hour and unblocked it the next day.
He says to find and attack the slowest part of the system. Output is limited by the slowest part, often IT, legal or procurement. Product managers are the new bottleneck because tooling went to engineers first.
The line comes from Fareed Mosavat. Speeding up one part of design, PM and engineering just moves the bottleneck.
7How Balfour hires, plans and structures teams at Reforge
Balfour hires on work samples and regrets every exception. At Reforge, 80 to 90% of a hiring decision rests on a portfolio review or a simulation. Conversations are weak and biased, a point he credits to Frank Slootman.
His team pushes back because simulations add work and slow the pipeline. He applies the same idea to promotions, with a light log of what each person helped ship.
He asks teams what it takes to win before the cost. Ideas rising from below get filtered by fear of rejection, so leaders see watered-down plans. He wants the full plan first, then trims it together.
"I don't want to be pushing the team. I want the team pushing me."
Reforge plans with visuals of the experience, not heavy OKRs. Planning processes during hypergrowth were disasters. Now 80 to 90% of planning assets are visuals of the experience, such as draft marketing assets, at about 75 people.
He keeps the end state directional and revisits it often. Whether this scales is a big question mark, since OKRs exist for scale reasons.
He says the year is made in the first six months. Plans that reach the target through a second-half inflection are unrealistic. SaaS buying windows run from a few months to a year, so late changes land in next year.
Teams also overestimate how fast a change shows up, and more so at scale.
Reforge flattened its org with captain IC roles. Every level has the same titles, there are more captain roles than manager roles, and Balfour plans to pay captains more than managers.
He wants to stop the 'manager death cycle', where star ICs leave the field and the new managers create meetings and process because they think that is the job.
He seeds bets with small teams and limited funding. HubSpot seeded several bets when it went multi-product, and only the CRM and sales tool survived. Reforge made 'small teams do bigger things' its second cultural value.
Reforge funded some half-working projects over the last two years and later unwound them. He says piling money and people onto them makes them worse.
8Balfour's personal habits and mindset
Balfour expects problems never to end. Solving one problem means taking on a bigger one. Reforge worked Ray Dalio's 'problem solving machine' passage into its cultural values.
He learned this the hard way and still rereads the passage when he catches himself hoping one fix will make life easy.
He keeps a doc of one-line lessons and rereads it when stuck. It holds a hundred and some entries, each a one-line lesson with a few sentences of description. He started it only a few years ago.
He used to take heavy book notes, but sifting through them added friction. Usually two or three lessons jump out and unblock a strategic question.
He aims to hand his sons more decisions each year. A parent's job, in his view, is to move the share of decisions made for a child down to zero by age 18. His sons are five and three.
His older son spends money saved from his grandparents and learns the consequences. In hard stretches at Reforge he returns to the 'man in the arena' quote.
Why ChatGPT will be the next big growth channel (and how to capitalize on it) | Brian Balfour
Platform shifts follow a four-step cycle (competition, moat, open platform, closed platform), and ChatGPT looks set to enter the open third-party platform step within about six months. The episode explains how to pick a platform bet, how to plan your exit before the platform closes, and how companies that set hard constraints are getting real AI adoption.
Balfour ran growth at HubSpot and co-founded three other startups. He was VP of Growth at HubSpot and co-founded three other startups before becoming founder and CEO of Reforge.
His lessons on growth, hiring and planning come from that mix of building growth at HubSpot and founding companies.
He lived a platform's glory days and horror days early on. In one of his first companies he built on Facebook's platform and saw both the boom and the crash when it closed.
He has also watched Google Ads, SEO and the Apple App Store rise and fall, which is why he treats platform shifts as a pattern.
Reforge began as courses and added software once AI made it possible. Learners kept asking for tools to implement what was taught. Reforge said no for a long time, until AI made it possible to encode that knowledge into software.
Reforge Insights now gathers feedback from every source, finds the gaps and auto-generates research to fill them. For ten years companies also came to Reforge to drive transformations, and AI adoption is the latest one.
Balfour says a great product is necessary but not sufficient. Winners are the teams that get distribution before incumbents can copy them. He has spent years pushing back on advice that boils down to product quality.
He calls this race for 'escape velocity' a good summary of what a startup does, an idea Alex Rampell wrote up around 2015.
The race for distribution is harder now than before. Incumbents copy faster, search clicks and social reach are shrinking, and AI code generation multiplies competitors in every category.
New platforms are one of the main ways out. Cursor overtook GitHub Copilot in nine months or less.
Every platform runs the same four steps. Step zero is a fierce market with five to seven players. Step one is finding the moat, step two is opening a third-party platform, and step three is closing it.
Closing means shutting the platform down, absorbing top use cases as first-party apps, or suppressing organic reach to push paid. Facebook, Google, iOS, LinkedIn and Udemy all followed it.
He sees the cycle as competition, not malice. Platforms must keep growing and must avoid enabling their own disruption, so they close. Sometimes closing is better for users, as Facebook's channels did get spammy.
His view: "They're playing you. So you need to play them."
The cycles are getting shorter, so there is less time to play. Facebook's cycle took about five years and Google's took many more. Facebook opened in 2007 at about a quarter of MySpace's size, then took a cut, suppressed organic reach and absorbed uses like events and photos.
Udemy started with about 80% of revenue going to creators and later cut that share sharply.
Balfour predicts winners by retention, not by reach. Facebook was smaller than MySpace and Friendster in 2007 but retained better. Google and Facebook won on retention and engagement, not on the biggest distribution at the time.
"it was never the person who had the biggest distribution at the moment of time. It was the one that had the best retention and engagement."
Balfour expects ChatGPT to open a third-party platform within about six months. He thinks AI has had a technology shift but not yet a distribution shift. A couple of years in, he sees all the ingredients for a new platform in place.
Casey Winters made the point that the two shifts rarely arrive together and the distribution shift comes a little later.
He argues the AI moat is context and memory, not the model. Leading models give about the same result, so the difference is which one holds more of your context. The more you use it, the more memory it stores, which feeds personalized outputs and another flywheel.
ChatGPT shipped memory early and keeps investing in data connectors, which he calls really context connectors.
Retention data backs his ChatGPT pick over Gemini's reach. Menlo Ventures data shows ChatGPT's retention curves level off higher than rivals' and form a rare smile curve. It has roughly 10x Claude's monthly users.
He reads the retention shift as the effect of memory. He wrote a post arguing against his own pick and says he may be wrong on ChatGPT but is confident the cycle will repeat.
He expects preferred partners first, then an open platform. The sequence he predicts is a launch with 10 to 20 brand-name partners that lend credibility, then a wider opening where the value exchange is defined.
Signals he cites include agent platform job postings, agent mode, shopping built into the interface and deeper attribution in results. He would bet a large portion of his net worth that OpenAI is already meeting potential partners.
He expects specialized agents because horizontal ones are hard to adopt. Users struggle with tools that do everything, and specific use cases need their own UI and data. That is why third-party agents should follow ChatGPT's general agent mode.
He also expects outcome-based agent pricing to erode as rivals undercut it, unless it is paired with a moat like proprietary data.
Balfour says never opt out of the platform game. Skipping it means competitors join and customer expectations shift to the new experiences. He calls it a prisoner's dilemma, and being early beats being late.
HubSpot joined ChatGPT's connectors even though exposing data looks irrational in isolation. ChatGPT now sends Lenny's newsletter more traffic than Twitter, so blocking it would only hand that share to someone else.
Startups should pick one platform and go all in. Late-stage companies can spread bets and wait for a winner. Startups lack the resources, and the failures tend to be those playing several platform games at once.
On mobile, betting only on Android probably lost, while iOS alone could still win. A winning strategy had to include the platform that won.
He picks platforms by retention, user quality, value exchange and scale. In that order. Scale decides only when the gap is huge, for example 200x.
Android has 70-some percent of devices but only about 30% of dollars, and iOS is the exact flip.
He says to plan your exit the day you enter. Build defensibility by owning a workflow, accumulating specialized data and context the big platform lacks, or building micro network effects.
You may not know the exit at entry, but figure it out along the way rather than waiting for step three.
He tells teams to be ready to turn strategy on a dime. New platforms must be exploited extremely quickly. The hard part is often leaders' reluctance to cause whiplash across projects already in flight.
Until the value exchange is known, he suggests working out how you would evaluate each player's platform and building relationships with the owners.
Balfour says hard constraints beat memos and incentives. One company capped each function at a fraction of peer team size for its revenue stage, so nobody could hire above it. Another rule is no new headcount until you prove AI can't do the work.
A smaller version is an executive who won't review a PRD without three prototypes. Clear messaging, owners, career ladders and review questions help, but he says constraints are the most impactful lever.
He sorts teams into catalysts, converts and anchors. Catalysts experiment on their own time, converts need structure, permission, budget and rewards, and anchors drag their feet and create silent friction.
Fewer than 10% of companies give anchors a date to change or leave, yet they are furthest along. His reasoning: "Cultures thrive on density."
He tells executives to audit adoption on the ground floor. About 90% of the time, a prototyping tool user said only they and one other person use it. Shopify measures real usage to stay close.
At one tech company a principal PM's prototype stalled for a month until the CEO heard about it at a happy hour and unblocked it the next day.
He says to find and attack the slowest part of the system. Output is limited by the slowest part, often IT, legal or procurement. Product managers are the new bottleneck because tooling went to engineers first.
The line comes from Fareed Mosavat. Speeding up one part of design, PM and engineering just moves the bottleneck.
He aims to hand his sons more decisions each year. A parent's job, in his view, is to move the share of decisions made for a child down to zero by age 18. His sons are five and three.
His older son spends money saved from his grandparents and learns the consequences. In hard stretches at Reforge he returns to the 'man in the arena' quote.
Brian Balfour: 10 lessons on career, growth, and life
Brian Balfour walks through lessons from a personal Notion doc of over a hundred, covering hiring on work samples, light planning, flat org design, and growth as a system. He also covers doing the opposite of the crowd, use cases over personas, and defining who you are not solving for, with examples from HubSpot and Reforge.
Balfour ran growth at HubSpot and co-founded three other startups. He was VP of Growth at HubSpot and co-founded three other startups before becoming founder and CEO of Reforge.
His lessons on growth, hiring and planning come from that mix of building growth at HubSpot and founding companies.
Balfour treats growth as one system, so causes sit elsewhere. Retention problems often come from acquiring the wrong users, and monetization problems from weak engagement. Teams skip asking which part of the system causes the issue.
Early HubSpot had weak renewals because sales sold to the wrong persona and use case. Removing commission on those deals changed behavior at once.
He defends good friction even when it cuts early conversion. The right friction improves downstream engagement and metrics, but short-term wins get applauded, so friction gets chipped away.
Reforge refined its application questions and turns away eager applicants likely to have a bad time and leave bad reviews.
He defines products by use case, then asks who has it. A use case is the problem, the value prop, the alternative and why they choose you. The natural frequency of the problem sets retention and activation metrics, and frequency of adoption sets acquisition.
Reforge calls this the use case map, developed with Casey Winters and Shaun Clowes. Students usually map about eight use cases, which is far too many.
He builds a relationship with buyers before they are in market. Most SaaS buyers are in market only once every few years. HubSpot's content made many people think of it as a blog, so it was the first place they went when ready.
He names who he is not solving for. Early HubSpot found four use cases hidden behind one marketing customer, chose only the mid-market all-in-one 'Marketing Mary', and aligned sales comp, success and marketing to turn the other three away.
The idea came home from his summers as a wedding planner in LA, where the most miserable couples tried to please everyone. Reforge's values doc spends half its length on what each value does not mean.
He studies what everyone does in order to do the opposite. When online courses were cheap, self-serve and open, Reforge launched high-priced, small-group, live and selective, and cohort-based courses became a thing afterward.
He applies it to ad creative too: review every competing advertiser's creative, then build something completely different. When copiers arrive, flip the pendulum again.
He wrote long, infrequent content while HubSpot's norm was high-volume. The playbook then was short articles, around ten a day. Most of his early email subscribers came from his very long posts.
He notes the HubSpot approach worked for HubSpot because of its early domain authority.
Balfour hires on work samples and regrets every exception. At Reforge, 80 to 90% of a hiring decision rests on a portfolio review or a simulation. Conversations are weak and biased, a point he credits to Frank Slootman.
His team pushes back because simulations add work and slow the pipeline. He applies the same idea to promotions, with a light log of what each person helped ship.
He asks teams what it takes to win before the cost. Ideas rising from below get filtered by fear of rejection, so leaders see watered-down plans. He wants the full plan first, then trims it together.
"I don't want to be pushing the team. I want the team pushing me."
Reforge plans with visuals of the experience, not heavy OKRs. Planning processes during hypergrowth were disasters. Now 80 to 90% of planning assets are visuals of the experience, such as draft marketing assets, at about 75 people.
He keeps the end state directional and revisits it often. Whether this scales is a big question mark, since OKRs exist for scale reasons.
He says the year is made in the first six months. Plans that reach the target through a second-half inflection are unrealistic. SaaS buying windows run from a few months to a year, so late changes land in next year.
Teams also overestimate how fast a change shows up, and more so at scale.
Reforge flattened its org with captain IC roles. Every level has the same titles, there are more captain roles than manager roles, and Balfour plans to pay captains more than managers.
He wants to stop the 'manager death cycle', where star ICs leave the field and the new managers create meetings and process because they think that is the job.
He seeds bets with small teams and limited funding. HubSpot seeded several bets when it went multi-product, and only the CRM and sales tool survived. Reforge made 'small teams do bigger things' its second cultural value.
Reforge funded some half-working projects over the last two years and later unwound them. He says piling money and people onto them makes them worse.
Balfour expects problems never to end. Solving one problem means taking on a bigger one. Reforge worked Ray Dalio's 'problem solving machine' passage into its cultural values.
He learned this the hard way and still rereads the passage when he catches himself hoping one fix will make life easy.
He keeps a doc of one-line lessons and rereads it when stuck. It holds a hundred and some entries, each a one-line lesson with a few sentences of description. He started it only a few years ago.
He used to take heavy book notes, but sifting through them added friction. Usually two or three lessons jump out and unblock a strategic question.