Dialing in an AI product's ICP and growing it to $400k ARR

Snow Lee is a 3x founder who is currently solving a problem he faced at his last company by building Runbear.

After struggling to find his ICP, he finally dialed it in and grew his company to $400k ARR.

Here's Snow on how he did it. 👇

A 3x founder solving his own pain points

I'm a 3x founder and have been building software startups for the past 16 years with my co-founder, Liam. I previously founded and sold two companies, and also led products as CPO at Buzzvil, where I helped grow annual revenue to $80M.

At my previous company, I led seven teams simultaneously. It was exciting, but exhausting. I was constantly context-switching, and the cognitive overload was brutal.

That experience made me realize how much time teams waste on repetitive communication. I wanted to build something that could help people offload that mental load, something that works with you, not for you.

That's why I’m building Runbear, a platform that helps non-technical teams create their own AI agents effortlessly. Our mission is to make AI feel like a real team member — one that can understand conversations, take initiative, and proactively help teams get work done.

We're currently at $400k ARR.

The challenge for finding your ICP

Our biggest challenge to date was defining our ideal customer profile. Since Runbear could technically help anyone build AI agents, we had to learn who got the most value.

Our first version was an AI copilot for DevOps engineers, but we quickly realized we were competing with our own customers — engineers wanted to build their own agents. So, we pivoted to helping teams utilize AI agents more easily.

Then, we discovered something surprising: Non-technical users — leaders, customer success, account managers, and operations teams — were the ones most eager to use our solution. The reason was that these communication-heavy roles benefited hugely from Runbear's ability to monitor messages and automatically take over tasks.

What surprised me most was how clear their intent already was. Even without understanding how AI works technically, our customers knew they needed it. They weren’t curious about the technology itself; they were focused on outcomes.

So, we refocused Runbear on helping non-technical teams harness AI effectively. And now, our strongest champions are team leaders, people who want to stay ahead and empower their teams through AI. And almost every customer starts with the same use case: answering repetitive questions.

It’s a simple but powerful entry point that quickly demonstrates value and builds trust in what AI can do for them.

If I had to start over, I’d focus on finding the killer use case first, before expanding the product’s capabilities.

How to find your ICP

To find our ICP, we started by analyzing existing use cases and identifying which customers were getting the most consistent value from Runbear. Then, we conducted in-depth customer interviews to understand their workflows, pain points, and expectations for AI assistance.

From there, we mapped out industry opportunities and prioritized problems that were either urgent, frequent, popular, or expensive, ideally hitting more than one of those criteria.

This framework helped us narrow down to customers who rely heavily on communication, such as account managers, support teams, and operations leaders, who now form our core ICP.

Boosting revenue per account

Until July, we had a single subscription plan, which helped us simplify growth early on. We had only one hard limit: the number of AI agents customers could create.

Many of our customers started reaching out to remove that restriction, which showed us a clear demand from power users. So, instead of just unlocking the limit, we decided to introduce a new business plan that better supported those teams, while charging more fairly for the added value. We wanted to test if enough customers were willing to pay more, and the strong response confirmed that they were, giving us confidence to move forward. We have already onboarded over 20 higher-paying customers, which has significantly boosted our revenue per account. There has been no churn from these users so far.

A tech stack that learns

Here's our tech stack:

• TypeScript

• Vercel

• Supabase

• Multiple AI models

• Claude: Core generation model

• OpenAI: Evaluation, knowledge extraction, and response compression

• And we support Gemini and Perplexity if users want them as their core generation model

Runbear’s biggest advantage is our team-member approach.

Instead of asking users to build complicated rule-based workflows, our AI learns how to work like a human teammate: observing, understanding, and acting naturally.

The agents analyze message intents to understand what’s happening in conversations, then decide whether to step in or stay silent, just like a thoughtful coworker would. When action is needed, the AI can plan the task completion process, execute steps, and even learn how to improve through ongoing conversations. Over time, it adapts to the team’s workflow naturally, extending its capability.

Although there are many ways that we ensure that AI acts as a teammate in this way, one example explains it best. When you ask the AI to do something it can’t do yet, it will ask you how to complete the task, just like a new hire would. You might say, "You can find the revenue data from Stripe," and it will learn and remember that instruction for the future. Over time, it builds knowledge across 2,000+ services, gradually becoming smarter and more capable, truly acting like a teammate who learns by doing.

That simplicity means anyone, not just engineers, can have AI agents that truly collaborate with their team.

Organic growth

We’ve grown mainly through content marketing and creator collaborations.

For content marketing, we mostly publish use cases, building guides, and customer success stories. For creator collabs, we publish youtube videos, X posts, and LinkedIn posts to introduce use cases of AI agents, and how to build them using Runbear.

Our SEO strategy has been effective, too. We’ve secured keyword positions that keep bringing in new users organically. To get there, we published content that was centered on keywords related to using AI agents in communication channels. Now, we're changing our focus slightly to keywords related to our ICPs and their use cases.

Success comes from staying in motion

Just start. Don’t overthink it. But once you start, keep evolving fast.

The market moves constantly, and your job as a founder is to adjust quickly. Build, learn, pivot — repeat.

Success comes from staying in motion.

What's next?

My life goal is to build a company that lasts 100 years. For Runbear, our vision is to become the de facto standard for hiring AI teammates.

In the short term, we’re focused on reaching $2M ARR and continuing to expand our customer base globally.

You can follow along on my personal blog. And check out Runbear!