Founder Playbook · Sub Club by RevenueCat
11 tactics from Ajay Mehta
Fueling Growth with AI and Viral Product Features — Ajay Mehta, Portola
Watch the full episode“Being forced because of high token costs to monetize relatively early on has actually been kind of a real help to our both growth and then also product development right cuz like we're not kind of wasting money on growth right now.”
Forced Monetization 2 Weeks After Launch Became A Product Clarity Forcing Function
Portola launched Tolen without monetization and within 2-3 weeks faced massive token bills as users chatted 30-40 minutes daily. Being forced to add a paywall immediately turned out to be a gift: paying customers send unambiguous signals about what to build next, and subscription renewal data is far cleaner feedback than free-user churn.
“We're hiring folks like it could be indie folks that live that are in college in New York City... they're just sort of like making a bunch of content and we're giving direction and like we have almost like a chat of folks that are making content together that are coming up with different ideas together.”
Build A Stable Of Part-Time Creators: Pay Per Post, Post From Their Accounts, Test 20 Angles
Portola's TikTok engine isn't an in-house team or an influencer agency — it's a networked stable of part-time creators who post from their own accounts. They share a group chat for brainstorming, each tries different content angles, and Portola pays per post. One in 20 videos goes viral. This is cheaper than big influencer deals and naturally tests more angles faster.
“You almost completed what is sort of like a voice AI personality quiz along with these beautiful animations... we tried to make it like something that would actually like questions that would be fun to answer about themselves and then actually give the user some output.”
Onboarding As A Voice AI Personality Quiz With Answers That Carry Into The First Conversation
Tolen's onboarding is a spoken personality quiz — fun, oddly-specific questions like 'what part of TikTok do you get stuck on?' — powered by voice AI with character animations. Crucially, the answers don't get discarded: they seed the Tolen's memory bank so the first real conversation references them. The paywall question then lands after the user has already felt the magic of personalization.
“Retention yeah I think has to be the north star because... we borrow from the kind of games world I know they they frequently look at that how many players are coming back on D7 D15 D30 it's just super useful for us.”
Borrow Retention Metrics From Gaming: D7, D15, D30 Are The Compass For An AI Companion
Ajay's north-star metric is not ARR or downloads — it's D7/D15/D30 retention borrowed from the games industry. For a companion app, the logic is airtight: if the product is truly valuable, each conversation should deepen the relationship and make the next one more likely. Hit-and-miss early retention means the product hasn't yet crossed the activation threshold for most users.
“Because the tolens are so bright and visual we were able to like have a relatively low cost per install from the get-go even on paid so it kind of gave us a lot of leeway to just experiment.”
Design For TikTok Virality From Day One — Visual Brightness Is A Distribution Feature
Portola didn't start with TikTok in mind but discovered that their deliberately bright, visual alien characters were naturally optimized for short-form video. The low cost-per-install on paid ads was a direct result of asset quality. Ajay's lesson: consumer app founders should evaluate distribution fit during character/design decisions, not after.
“Raise as little as you need to validate an idea right and then once you have progressively... set expectations very low like low you know mid six figures like just enough for you and your co-founders to not totally like starve.”
Raise The Minimum Viable Pre-Seed — Just Enough To Eat While You Wander The Idea Maze
Portola's early funding strategy: a mid-six-figure pre-seed from one trusted investor, purely to keep co-founders alive during exploration. No product, no deck — just an idea and a credible team. Once the soft launch showed real traction, the same investor doubled down. Ajay's point: naming rounds (pre-seed, seed, Series A) obscures the simple truth — you're selling future cash flow promises at a clearing price.
“Most people do talk to their tolen via voice and we tried to make that like the best quality voices possible the fastest response times possible the best memory system... once you actually start trying to build a awesome voice AI experience there's a reason that there are like two or three on the app store.”
Voice AI With Memory Is The Core Technical Defensibility — Most Competitors Won't Replicate It
Tolen's real defensibility isn't the cute alien art — it's the engineering underneath: low-latency voice AI, parallel prompt streams for conversation classification, and a memory bank that resurfaces personal details across sessions. Ajay notes that while competitors can copy the visual design, replicating the actual voice + memory experience is what keeps the category sparse.
“Finding the way you can get scalable installs as cheaply as possible is part of your if not your product market fit your product market channel fit right like figuring out how you can actually like make this a scalable like situation.”
Find Product-Market Channel Fit: Scalable Organic Distribution Is Part Of The Product
Ajay distinguishes PMF from what he calls product-market-channel fit — knowing not just that users want the product, but that you can reach them cheaply enough to build a sustainable business. For Tolen, TikTok virality (from visual brightness + relatable content angles) drove cost-per-installs so low that paid ads became nearly redundant. Channel fit should be evaluated alongside product fit.
“Before LLMs and like being able to just like stuff context into a machine... you have to take that general purpose technology and multiply it against every previously existing niche.”
AI Enables Hyper-Personalized Onboarding That Answers All Those Questions Surveys Used To Waste
Portola applies a key insight to onboarding: every question you ask a user in a survey can now be fed directly into an LLM context window and actually used. The combinatorial explosion that made personalized onboarding impossible to ship procedurally is now trivially handled. This makes the investment in a long personality-quiz-style onboarding actually worth it — the model will use every answer.
“The most compelling part of what we had been building were like highly personalized experiences... something that AI really allows you to do is like make consumer experience that just feels like there is totally something on the other side that knows you understands you.”
Emotional Resonance Is The Insight That Separates Retained Users From Bouncers
Ajay's founding insight wasn't a feature — it was an emotional reaction. Early prototypes that felt truly personalized produced a categorically different user response. That 'warm' feeling became the product north star. For retention in a subscription app, emotional investment outperforms utility: users who feel known by a product don't churn, they evangelize.
“A woman posted a TikTok we have no affiliation with just talking to her tolen about her favorite WWE wrestler that's like picked up like 30-40,000 views.”
UGC Starts Happening Naturally Once You Have Enough Brand Affinity — You Cannot Buy That
Portola hit 800K+ downloads partly through a creator network — but the signal Ajay values most is unaffiliated UGC: users posting about their Tolen with zero financial incentive. A woman talking to her alien about WWE wrestling picked up 40K views. This organic content is both validation of real product love and the beginning of a brand moat that money can't replicate directly.