Apt

Apt started with one rigorous assessment that scores many personality tests at once. A single 15 to 20 minute assessment produces scores on Big Five, Holland Code, MBTI, Jung archetypes, DISC and Enneagram, which Apt pitches as "one test to rule them all." About two million people have taken it.

The hard part was choosing questions so one sitting could feed every test, so Apt built its own algorithm and scoring to keep results valid, then added AI on top. The team started from hundreds of hours of assessments and kept only the scientifically valid, reliable ones.

Subversive▶ 20:00

The quiz built a proprietary dataset and a built-in way to spread. Two million completed assessments give Apt data on how people are wired across several hundred psychometric traits, which general chatbots do not compute. Host Phil Carter notes people like sharing personality results, so the quiz carries a viral loop.

Apt leans on existing research linking traits to career paths so its matches are predictive rather than guesswork. Tomas Hoyos calls it the gap between knowing yourself and choosing a career.

Subversive▶ 48:58

Expand from the wedge into the full career workflow. After scores and reports, Apt added career path recommendations, job matching and a 24/7 AI mentor that helps with applying, negotiating and thriving in a role, with a marketplace planned on top.

Each career path is explained: who thrives, a typical day, growth or decline, pay and exit options. The idea came from cofounder James, who watched his father's recruiting firm give candidates personality and aptitude tests; Tomas saw that AI could make the test a conversation.

Subversive▶ 20:00

Switching to a conversational AI quiz lifted onboarding conversion over 60%. Apt replaced multiple choice questions with a chat where the mentor reacts to each answer and asks a follow-up. Onboarding conversion rose by more than 60%.

It feels more enjoyable than a school-style quiz, and the probing questions make users believe the product understands them, a belief that carries to the paywall. No baseline or test details were given. Tomas: "it's so engaging and just drives up conversions so much that it's definitely worth it."

Subversive▶ 36:24

A longer quiz converted better because effort builds curiosity and trust. In Apt's A/B tests the roughly 15 to 20 minute quiz outperformed shorter versions.

Users who invest time are more curious about the results, and a deep set of questions makes the output more believable. Tomas: "when people are investing in something and taking the time to complete something, they're extremely curious about the results."

Subversive▶ 34:00

Use LLM analysis on free-form answers to recover segmentation. A conversational quiz loses clean multiple choice segments, so Apt runs LLM analysis on the conversations afterward to tag and bucket users and infer motivations they never stated.

It adds a step but opens more ways to personalize. Tomas: "the fact that it's probabilistic is like a feature, not a bug."

Subversive▶ 37:58

Instrument every step of the quiz funnel and tune each one. Track landing page visits, quiz starts, drop-off at each question, paywall views and paid conversions, and A/B test the copy that gets people to start.

Subversive▶ 34:57

Defensibility rests on proprietary data, workflow depth and network effects. Citing Bill Gurley, Tomas frames app-layer defense as three things: Apt's psychometric scores from two million assessments, depth across the whole career workflow, and a future marketplace with a data flywheel.

He expects every chatbot rival to prompt "why not ChatGPT or Claude?", much as founders were once asked "what if Google does this?"

Subversive▶ 48:58

A dedicated jobs database beats a chatbot's slow web search. Apt keeps over 10 million open jobs in its own database, splits each user profile into about 500 vectors and matches them to jobs with cosine similarity, then applies filters.

Tomas argues the AI labs are unlikely to care enough to improve job search through web browsing. No evidence was given that the matching predicts good hires.

Subversive▶ 52:44

High-stakes domains need an opinionated interface, not a blank chat box. Apt tailors its interface, messaging and niche features to deliver dense career information in an engaging way, and aims for a brand people trust and recommend.

Subversive▶ 53:40

Apt keeps betting on freemium despite it hurting conversion in tests. Apt charges from day one, and its freemium experiments so far lowered paid conversion versus a hard paywall. Tomas still favors freemium for reach and category creation.

His argument: new categories sound odd at first, as early Instagram did, so going free and trying to "blow it out" maximizes reach and could crowd out rivals. A large free job seeker pool would supply liquidity, and Apt would earn placement fees from employers instead of charging users $20 or $30 a month.

Subversive▶ 41:30

Design the marketplace to learn from every match outcome. Apt plans to match seekers to employers using deep profiles, then use the ongoing mentor relationship and employer feedback to judge each match and improve the algorithm.

More seekers would draw more employers and vice versa. Tomas calls it "a flywheel that gets smarter with every crank."

Subversive▶ 28:20

Calling it a mentor avoids coaching's stigma, cost and scheduling friction. Apt deliberately says career mentor, not coach, targeting a roughly $7 billion coaching market with something always available for less than a cup of coffee a day.

People skip coaching because trusted coaches are hard to find, it carries stigma, costs thousands and needs scheduled hours. The mentor calls tools for local pay data, open roles, contacts at target firms and resume fit for applicant tracking systems.

Subversive▶ 24:37

A proactive mentor turns a one-time tool into a habit. Apt wants the mentor to act unprompted: spotting below-market pay, building a 6 to 8 week plan to ask for a raise, and surfacing growing companies hiring for the user's profile with contacts and a cold email.

Subversive▶ 26:22

Shipping one experiment a week found a hit and a cofounder. After Tomas's health network Voro lost about 90% of monthly active users in two weeks when COVID hit in March 2020, his team launched a new product weekly. A Zoom live-event tool for creators, built in 48 hours, took $100,000 in gross sales its first week and grew to a few million in volume.

Voro had reached several million users organically. The events product faded as the world reopened, so the team sold assets and returned capital; James, an angel investor in it, became Tomas's cofounder at Apt in 2023.

Subversive