The Invisible AI: Growing from 0 to 3K WAUs by Making AI Disappear
Lessons from building Apollo's AI Power-ups from and why the future of AI products isn't about AI at all.
Everyone's rushing to put AI front and center in their AI products. The playbook seems simple: Integrate with a LLM, add a prominent AI button, let users craft prompts, and ~walla~
We tried that. It failed spectacularly.
Our path to 3,000 weekly active users in 3 months came from doing the opposite: making the AI disappear entirely. This isn't just another product story – it's a glimpse into how AI products actually succeed when we stop obsessing over the AI itself.
Part 1: The Beta - Start Small, Learn Fast
The key to launching AI products? Just like any other zero-to-one product: Talk to users and build a product at least one person LOVES.
We handpicked 10 beta customers with two specific criteria:
Clear need: Already using AI for sales use cases (e.g. conducting research, targeting leads, crafting messaging) but struggles to do so today inside Apollo
Eager to give feedback: Willing to engage deeply in giving feedback on AI Power-ups, as evidenced by their recruitment surveys and recent LinkedIn activity
With each of our beta customers, we hosted dedicated “jam sessions” where we would help our customers live on a call build out custom AI power-ups for their unique use cases. These meetings weren't your typical “show & tell” onboarding calls – they were working sessions that helped us build deep customer intuition. This is similar to what Clay calls “reverse demos” in their sales process.
This structured approach paid off:
60% success rate with beta partners
2/3 of users reported they were likely or very likely to use power-ups again
Majority of prompts rated 4-5 stars on tone, phrasing, and natural language
But the real gold came from watching users struggle. We saw them:
Spending 15+ minutes configuring an AI Power-up
Getting stuck on model selection
Struggling with prompt engineering
Unsure how to leverage Power-ups in their existing workflows on Apollo
By helping our beta customers unlock their first “aha” moment with AI Power-ups live on a call, we unlocked key nuggets of customer insights that informed our future roadmap, but even importantly, we formed real partnerships with our beta customers.
In other words, we found a group of partners who LOVE AI Power-ups. One such partner was Grace from Smartling. Her team was able to 10x their personalized email volume and save 2-3 hours a week of manual research per SDR.
Part 2: The Launch - Big Bang vs. Steady Growth
Once we built the first version of a product someone LOVES, the next thing we wanted to do was tell everyone about it.
So, we did the big launch - hosting a major GTM moment with Apollo AI Power-ups & Workflows in late October that drove our first growth curve, growing WAUs 7.7X to 1,000 in one week. Press releases went out. LinkedIn posts got engagement. The classic playbook worked... initially.
But the sustained growth came from what happened next. Instead of chasing the next big feature announcement, we focused on:
Consistently sharing customer success stories
Building relationships in the space
Regular updates showcasing real user outcomes
A clear insight emerged: The best marketing wasn't about AI capabilities - it was about customer outcomes. Seems obvious, but as a technologist, it’s easy for us to get excited over the technology instead of actually solving customer problems.
We shared Grace’s story of how her SDR team used AI Power-ups to identify website translation gaps. Their outreach went from generic personalization ("I see you like Guinness!") to value-driven insights: "We noticed these specific gaps in your website's Spanish translation. Here's how to fix them..." Read more here.
When we shared how Grace at Smartling used AI Power-ups to identify website translation gaps, it resonated far more with our customers than any feature announcement. Customers want help solving their own problems - not more features.
Part 3: The Breakthrough - Create Magic, Hide Complexity
However, the real breakthrough came when we finally solved our customer’s #1 pain point when using AI Power-ups: writing a prompt. Many of our customers found it challenging to pick the right model, craft the right prompt, and reference the right variables to get the output they want - so many gave up.
Inspired by Google’s core insight that “the user is never wrong,” we decided to re-think our prompting interface entirely where our users could simply state what they want to do, and get results. We would handle all of the back-end complexity for our users. As Aravind Srinivas, the Perplexity founder, put it, “it’s our job to create magic behind the scenes for our users. Let our users be lazy.”
The technical challenge was significant. Our Assisted prompting interface needed to:
Auto-detect the most relevant model for each use case
Generate contextual prompts based on user intent
Handle variable injection using Apollo fields
Format outputs for immediate use
Early users were spending up to 15 minutes configuring each Power-up. With Assisted Prompting, that dropped to under 60 seconds. The data told the story:
3x jump in WAUs since releasing this feature
32% improvement in enrichment completion
A small but telling example: early users kept asking "which model should I pick for this use case?" We realized that was the wrong question entirely. Users shouldn't have to think about models any more than they think about database optimization when using Google. It’s our job to create a magical experience for our users.
Part 4: The Final Piece - Platform Integration
Here's what everyone is waking up to in the AI industry: the magic isn't in the model, it's in integrating AI seamlessly into user workflows.
The game-changer for AI Power-ups was integrating our AI functionality across the Apollo app, so wherever our users go, they can leverage the data they gathered from AI power-ups for any use case, such as:
Running AI Power-ups in bulk across any list
Applying filters on prospects using data points gathered using AI Power-ups
Dynamically inserting these AI Power-ups to personalize any message
This "create once, use everywhere" approach for AI Power-ups drove a 23% jump in retention in Q4. Why? Because users were able to form a habit around using AI Power-ups for any workflow inside our platform.
The Future of AI Products
Three key lessons emerged from our journey to 3K weekly users:
The "AI curiosity" phase is over. Users don't want to spend time prompt engineering or exploring models. They want results fast.
Integration beats isolation. The most successful AI features aren't standalone tools – they are deeply woven into existing workflows.
Clear beats clever. Every time we made AI less visible and outcomes more visible, usage increased.
The metrics tell the final story:
0 to 3K WAUs
+28% lift in email sent to interest rates
17.5% free-to-paid conversion (9.3x above baseline)
Looking ahead, I believe the next wave of successful AI products won't be marketed as AI products at all. They'll be tools that just work better, faster, and smarter – with AI humming quietly in the background.
The future of AI isn't about making AI more visible. It's about making it disappear entirely.