Aug 14, 2026 · 1h 20m · news
How to Build a $100M Growth Engine: Lessons from Wispr Flow & Superhuman | Matt Swulinski
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, growth leader Matt Swulinski joins Harry Stebbings to explain how SaaS startups can build $100M growth engines by combining direct-to-consumer performance marketing playbooks, robust data tracking, and AI agent automation.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 17.9% of the talking time here. How this is scored →
Harry as informed peer 5.9 Guest teaching 6.4 Guest disagreement 2.7 Harry pushing back 2.7
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 38:48 Defending Victor's broad positioning against Stebbings' attack
Swulinski firmly pushes back against Stebbings' blunt rejection of Victor's tagline, defending horizontal positioning as necessary to build early category awareness.
Hardest push from Harry ▶ 38:48 Stebbings challenges Victor's vague AI employee taglineStebbings directly refuses Swulinski's framing, arguing that calling a product 'the AI employee for everyone' is confusing compared to vertical positioning.
Biggest teaching moment ▶ 58:30 Architecture of an automated Claude Code marketing agentSwulinski delivers a masterclass on how he built an autonomous Claude Code system to manage 70 to 120 newsletter partnerships end-to-end without manual media buying.
Harry holds his own ▶ 1:16:49 Stebbings details 20VC's proprietary AI investment graderStebbings demonstrates deep operational AI implementation by revealing how his investment committee automated deal stack-ranking across over 1,000 startup calls.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Adapting Product-Led Growth for the AI Agent Era | 5 | 6 | 2 | 2 | Stebbings introduces Swulinski and asks how the rise of AI agents alters product-led growth dynamics. Swulinski outlines why the e-commerce growth playbook of aggressive paid validation and UGC creator networks applies directly to modern SaaS. | |
| Validating PLG and Core Paid Channels | 5 | 6 | 3 | 3 | Stebbings notes common venture advice warning founders away from paid acquisition early on. Swulinski rejects the advice, arguing paid ads are the fastest way to validate messaging, funnels, and true PLG viability across Meta and Google. | |
| Building Modern Conversion Tracking and MarTech Architecture | 5 | 7 | 2 | 2 | Swulinski explains the deep MarTech divide between e-commerce and SaaS, noting the lack of out-of-the-box multi-touch attribution tools for software startups. He details the necessity of custom conversion tracking and match rate optimization before spending. | |
| KPI Targets, Acquisition Ratios, and Token Economics | 6 | 6 | 2 | 3 | Stebbings probes early CAC:LTV economics and tolerance for high burn. Swulinski breaks down the transition from upper-funnel volume metrics to fully loaded gross profit modeling factoring in heavy LLM inference and token costs. | |
| Scaling Creative Production and Creator Incentive Programs | 5 | 7 | 2 | 2 | Swulinski explains how Meta's Andromeda algorithm transformed creative into targeting, necessitating 400 to 500 net new creatives monthly. He describes setting up revenue-share UGC creator incentive programs to sustain that creative throughput. | |
| Pattern Disruption, AI Content Limits, and Testing Windows | 6 | 6 | 2 | 2 | Stebbings questions video hook mechanics and the viability of fully synthetic AI creative. Swulinski agrees that full AI videos are mostly unconvincing slop and emphasizes pattern-disrupting human UGC across diverse creator demographics. | |
| Web Optimization, Page Speed, and Positioning | 6 | 6 | 3 | 3 | Stebbings pokes fun at unconventional UI designs like PostHog. Swulinski defends deliberate pattern disruption for specific technical audiences while highlighting the critical impact of mobile UX and page speed on conversion and AEO crawler indexing. | |
| Measuring Incrementality, Spend Elasticity, and ICP Expansion | 6 | 6 | 4 | 4 | Stebbings asks why Superhuman's initial growth curve hit an asymptote and questions if they mismanaged paid growth relative to competitors. Swulinski explains the limits of single-ICP founder targeting and the necessity of testing spend elasticity. | |
| Enterprise PLG, YouTube Ad Formats, and Execution Mechanics | 5 | 7 | 2 | 2 | Swulinski outlines how enterprise expansion happens organically through bottom-up PLG adoption. He also shares a tactical hack for converting 9:16 vertical creator clips into landscape 16:9 YouTube video ad assets via automated framing. | |
| Channel Saturation, ICP-Specific Marketing, and Product Taglines | 7 | 5 | 5 | 7 | Stebbings delivers sharp pushback against Victor's broad tagline ('the AI employee for everyone'), calling it vague and uncompelling. Swulinski defends the horizontal positioning as necessary to seed initial category awareness before verticalizing into niche workflows. | |
| Viral Referral Architecture and Paywall Mechanics | 6 | 7 | 2 | 2 | Swulinski outlines referral design, advocating that prompts be triggered right at usage limits rather than gamifying complex tiers. He explains Victor's credit-based B2B referral system where users receive rev-share payouts in tool credits. | |
| Answer Engine Optimization (AEO), YouTube Reviews, and Modern PR | 6 | 7 | 2 | 2 | Stebbings asks about the utility of Answer Engine Optimization (AEO) and whether traditional PR is dead. Swulinski explains that modern PR and long-form YouTube reviews serve primarily as authoritative citation sources ingested by AI search engines. | |
| Paid Acquisition vs. Organic Growth Mix & Incrementality Testing | 7 | 6 | 2 | 2 | Both host and guest strongly agree that modern growth demands full-stack generalists capable of end-to-end execution rather than siloed specialists. Swulinski recounts running multi-million dollar ad budgets solo at Whisper through deep operational fluency. | |
| Evolution of Growth Talent: AI Natives & Systems Thinkers | 6 | 7 | 3 | 2 | Swulinski argues that corporate attempts to make teams AI-native are failing because employees lack systems thinking. He defines systems thinkers as operators who deconstruct entire workflows into modular inputs and automated feedback loops. | |
| Testing AI Workflows & Building Agentic Marketing Operating Systems | 5 | 8 | 2 | 2 | Swulinski details how he engineered an autonomous marketing operating system using Claude Code to handle inbound sponsorship negotiation, contract ingestion, UTM tracking, and performance analysis with minimal human touchpoints. | |
| Lean Agentic Teams, Future Organizational Models, and Restructuring | 6 | 7 | 5 | 4 | Swulinski delivers a provocative take that founders should fire non-systems-thinking marketing staff as teams shrink to hyper-leveraged solo operators. Stebbings challenges whether traditional enterprise CMOs have any real grasp of these agentic shifts. | |
| Blurring B2B/B2C Boundaries & Hiring Growth 'Unicorns' | 6 | 6 | 4 | 3 | Swulinski highlights external affiliate revenue-share programs as the most underrated growth channel and calls out X (Twitter) ads as totally ineffective for SaaS performance. Stebbings and Swulinski discuss elite viral talent like Tobin. | |
| The Proposal for an AI Execution Boot Camp | 8 | 6 | 2 | 2 | Stebbings proposes a dedicated execution bootcamp for AI operators and shares his venture fund's proprietary AI call-grading system. Swulinski shares his Obsidian session-end logging architecture for compounding agent memory. |
Statements from this episode (38)
Top PLG companies are best positioned for autonomous AI agent software selection
“So the best companies in PLG focus right now will also be best positioned, in my opinion, for the agentic layer of essentially decisions being made without that human in the loop when tools are being selected for certain tasks.”
E-commerce performance playbooks are the correct B2B SaaS growth strategy
“My philosophy is that the Ecom playbook
Is the right playbook for SAS?
Because if you look at e-comm, you have every single cent needs to equal a purchase or an ad to cart.
You have hundreds of UGC creators, variety of creative, and you also have a ton of chan…”
Swulinski: Paid ads are the fastest way to validate product-led growth
“Paid is the easiest way to validate that you have PLG, that you have a product that can scale in any way, shape or form, right? Like there, there's a ton of narrative that like focus on brand, go organic content, build all that base. That just takes too much t…”
Startups can scale to $10M ARR using only Meta, Google, and lifecycle
“You want those three things spun up because you can scale to your first 1,000,010 million ARR, Just off of those three things. And then obviously, you know, you're investing into the site, you're investing into founder-led content. There's like the other thing…”
Swulinski: SaaS Lacks Out-of-the-Box Multi-Touch Attribution and Conversion Tracking Platforms
“For SaaS, Completely different world because you have to have an engineering team that builds this from scratch, right? When it comes to analytics, there's a DB, there's a BI layer. So maybe you're using click house and hex last click, you know, first click at…”
Swulinski: 90% of Companies Fail to Build Proper Tracking Before Spending
“I'd say 90% of companies don't do that as a first step. Before you spend your first cent, have everything set up.”
Swulinski: Paid Acquisition Usually Fails Due to Bad Conversion Tracking Setup
“If you have poor conversion tracking, Meta doesn't know who those people are, and it just randomly targets people, and you, you'll have a super high CAC, you'll say, ah, pay doesn't work for me. I'd say most of the time, people haven't done the actual setup co…”
Swulinski: AI SaaS must calculate LTV to CAC on gross profit
“LTV to CAC is usually not enough. You want to probably look at like LTV gross profit, specifically if you're, you have token cost because you're, the biggest cost item is, is usage at the end of the day.”
Swulinski: Meta and Google Ad Algorithms Dislike Usage-Based SaaS Models
“The biggest differences are first in that, like, Meta and Google are not optimized for, like, B to B SaaS usage-based products. Like, they go absolutely nuts with a, so you're telling me there's a predictive revenue of a one person could be a 50,000, you know,…”
Swulinski: A $100K Meta Budget Requires 400-500 New Creatives Monthly
“Like if you, let's say you have a hundred K meta budget, You probably need at least 400 to 500 new creatives a month. Otherwise, you're going to plateau. You're going to get out-competed.”
Swulinski pays teenage UGC creators up to $30,000 monthly for ads
“Like, we have we have kids that are, like, 1718, 19, that are making 20, 30 K a month, just making a couple ads for us.”
Swulinski: Scaling only winning creatives causes ad performance to crater
“The key thing around creative is not the one creative. It's the package of all of your assets together. They have to be different. They have to, they, to the algorithm, they have to look and feel different. Different ages, different genders, different settings…”
Swulinski: AI video tools cannot replace real people in creative ads
“In no way, shape, or form is anything in creative to the point where you can replace real people.”
Swulinski: Startups need 50 conversions before running paid ad campaigns
“If that's already happening, then your conversion tracking is working on your website or your app or whatever, and you get to 50 conversions, right? Then you can start paid because that's like the minimum line of whatever your conversion action is that the alg…”
Swulinski: PostHog's unconventional website UI succeeds as audience pattern disruption
“That's like a pattern disrupt thing for them, and they know who their target audience is, and the target audience appreciates the weirdness.”
Swulinski: Superhuman hit a growth ceiling by hyper-focusing on tech founders
“At the end of the day, it was very much a, you know, premium product, immediate paywall, and the core ICP is founders, right? And we also experienced this at Whisper, right? Like the early adopter tech founder is not an infinite audience. I think today it's an…”
Swulinski: Large Share of Wispr Flow ARR Was Enterprise Before Hiring AEs
“We were B to C, PLG, but like a very large component of our ARR was enterprise before we had a single AE, because it was self-service, enterprise adoption, team licenses, and that kind of stuff”
Swulinski: Growth teams should spend aggressively to find channel ceilings
“What I think everyone should do, is go as hard as you absolutely can, and then see it blow up, and then understand where and how it blew up, pull back, and use that as information of now we know what is incremental, we know where our ceilings are”
Swulinski: Google Ads was Wispr Flow's best-performing acquisition channel
“Google Ads, it was, it is still the main driver, just and it was surprising across literally everything, right? Non-branded search, Pmax, YouTube ads, everything prints there”
Stebbings: 'The AI employee for everyone' is a bad tagline for Viktor
“As an investor in the company, I don't think that was a good tagline, and you may hate me for that, but the AI employee for everyone, I don't know what the fuck that means.”
Swulinski: Eight-person teams spend $15k-$20k monthly on Viktor
“We have like eight person teams that are spending 15, 20,000 dollars a month to use Victor. And it replaced all their hiring.”
Swulinski: Place SaaS paywalls inside the post-aha honeymoon phase
“You have that magical aha moment, And then you're kind of, you stay in that kind of like honeymoon phase of like, this is just amazing. And then within that phase, paywall, right?”
Swulinski: Free trial credits must be included in marketing CAC
“Our free trial credits get summed with marketing spend. If you're not doing that, then you're not really calculating your acquisition cost. So yeah trial credits or like anything free that you're giving away is cost, and it goes under, it is a marketing cost i…”
Swulinski: YouTube reviews are top citation sources for ChatGPT and AEO
“YouTube reviews, because they're long form, and they rank long tail. They'll essentially beep, they're one of the main things that get picked up by the answer engine. So you'll have your website, PR, external sources talking about you, and one of the biggest, …”
Swulinski: PR now primarily drives AI answer engine citations, not traffic spikes
“Really, that PR is made for citations, right, around AEO. Like, how many external credible sources are writing about you? Then you'll see your AEO traffic significantly increase, because the more of a not-owned narrative is being talked about positively about …”
Swulinski: TikTok marketing has not worked at any of his three companies
“To be honest, in any of the three kind of core companies that I've been at so far, It hasn't worked yet, but maybe I'll get a chance to change that at Victor.”
Swulinski: Healthy organic acquisition mix is 35% to 45%
“The sweet spot for organic mix should be around like, 35 to 45% of acquisition is word of mouth, organic, and like truly people just discovering it based on everything else that you've built.”
Swulinski: Spending over $1M monthly requires MMM and incrementality models
“You definitely want a strong MMM model and incrementality when you're, like, the moment you start spending north of a million dollars a month”
Swulinski ran $1M/mo Google spend and 100+ newsletters solo at Wispr
“For essentially a year and a half, I was the one person at Whisper doing all execution. So imagine a million a month on Google, hundreds of thousands of keywords, ads, experiments, landing pages. I owned every KPI from eyeball to download and then product tick…”
Swulinski predicts AI agents will execute 80% of company work in three years
“I think in three years, companies will essentially be, like, board of directors, where 20% of the work is the strategy and the thinking, And then agents just do 80% of the execution, right?”
Swulinski: Fire marketing team members who are not systems thinkers
“My hot take was probably fire most of your marketing team that is not a systems thinker. Like, stop brute forcing people in. Hire the right people into the roles because the role has changed.”
Swulinski: Marketing Acquisition Talent May See ML-Level Hiring Boom Within a Year
“I think we're coming to the place that like people in these marketing roles also are They're unicorns. They're rare, and you want them in your company, and I think that same like ML crazy moment is maybe going to happen over the next year in, in the marketing …”
Swulinski: 10% to 15% of Viktor acquisition comes from affiliate program
“Like for Victor, a good 10 to 15% of our acquisition on a monthly basis comes through our affiliate program. And that's, you know, people that want to earn money because we give them between 10 and 15% revenue share.”
Swulinski: Top affiliates at Viktor earn $20K to $30K per month
“We have some affiliates that are making 20, 30 K a month.”
Swulinski: X ads do not work for SaaS growth
“X ads don't work. And, like, that to me just means the platform is still... I have yet to meet a SaaS head of growth or performance marketer that says that X ads print.”
Stebbings: A 6-week AI bootcamp is more valuable than university
“Come to our school, and for six weeks, we will teach you how to get jacked up
On every system, process, model, for six weeks, nine to five, this is your job, and you pay 10,000 bucks.
I would say that's more valuable than university, say, for teenagers.”