Fueling Growth with AI and Viral Product Features — Ajay Mehta, Portola

Onboarding is framed as a story: every person has a unique Tolan on planet Portola, and to find yours you complete a voice-AI personality quiz with an animated 'Oracle' character. Questions are deliberately fun, like an Enneagram or astrology reading ('what oddly specific part of TikTok do you get stuck on?'), not functional ('what workout do you like?'). The app gives an output ('here's what we learned about you') and matches a Tolan. The quiz context carries into the first chats and is stored as memories, so the Tolan later says things like 'the Oracle told me you are really into cooking pizza right now'.

After TestFlight, put Tolan in the App Store as a bare-bones soft launch with no monetization, using the best models and voice quality available, only a basic usage limit, to see how real users behaved.

Because each long voice call runs several prompts at once (reply, conversation classification, memory lookup) across OpenAI, Anthropic and voice providers, heavy users became expensive. Two to three weeks after the App Store launch the team added a paywall and signed up for RevenueCat. Paid users then forced product improvements, and month-2/4/6 renewals became the signal for product priorities.

Hire non-professional creators (side gig, e.g. college students in New York or locals) to produce lots of short videos showing Tolan. Keep them in a shared group chat to brainstorm and try angles together, with direction from the team. Pay per video or on retainer, not per view. Videos go out on creators' personal TikTok/Reels accounts, separate accounts and the official Tolan account. Most get few views; winners are sometimes, but not usually, boosted with paid spend.

Base video angles on how real users use Tolan: studying late for an exam ('Up at 3:00 a.m. studying with my Tolan'), nursing-school revision, cooking, doing makeup, picking outfits, and relationship advice ('I'm at the talking stage with this guy in my class'). Keep trying new activities and see which catch in the algorithm.

Collaborate with influencers who already have large audiences (around 1M followers) to make a video posted on their own account, priced off their average views. Newer for the team; they plan to do more.

Worked with outside artists they admired online to design and animate the Tolan characters and onboarding; render the Tolan in real time in a Unity canvas inside the native iOS app. Iterated on details (eyebrows, how human-like, starting as a baby vs adult). Chose a cute, non-humanoid, non-anime alien to keep it friendly.

Deliberately designed the companion as a cute, cuddly, non-sexual alien and avoided humanoid or anime characters, pitching it as a friend that won't try to make you fall in love with it.

Over a few months after the paywall, improved conversation quality, animations and onboarding, and added a planet where the Tolan walks around as a shared space that grows with the relationship. The goal was to get more users to the activation moment (their first really long conversation). Tracks D7, D15 and D30 retention as the North Star, borrowed from games.

Raised a pre-seed from a known investor (Lachy Groom, ex-Stripe solo GP) on the three co-founders and a general direction while prototyping. After the soft launch showed traction, the same lead investor doubled down, making a $10M seed in total.

Alongside organic creator content, ran small-scale paid ads from the start just to bring people through the funnel and judge retention and conversion. Kept this setup for a few months; paid creative has its own separate process.