Hitting $500k ARR in two years by staying close to her customers

Tham (Sylvia) Nguyen saw a hole in the AI market and built a tool in a weekend to fill it. Then, she listened to her customers and created something entirely different.

Less than two years later, MindPal is bringing in $500k ARR.

Here's Tham on how she did it. 👇

Building it in a weekend

My cofounder and I started MindPal in August 2023 when we were still on summer break at college. Back then, we saw a lot of simple 'chat with your PDF' apps popping up. They were interesting, but my cofounder and I quickly realized they were quite limited. People work with all sorts of information – Word docs, websites, presentations, videos, etc. It's not just PDFs. So we quickly spun up the first version of MindPal, a place to dump all your files — any file — and chat with them.

We were able to build it over a weekend, partly because it was really simple — just a one-page app with a payment gateway — and partly because of the mindset and technical skills we gained after working on our previous failed startup. 

Before MindPal, my cofounder and I spent nearly 1.5 years on another startup. It failed because we were completely out of touch with potential customers. That experience taught us invaluable lessons. Most importantly, we learned what really matters when it comes to doing business: Building products people actually want. Nothing else matters if you do not have customers to serve.

We launched on Product Hunt right away. To our surprise, it ended up at Top 4 Product of the Day. Thanks to that launch, we got around 100 paying customers over the next 2-3 weeks.

We decided to stop going to school to work on MindPal full-time. We did get a small pre-seed check from a US VC a few months later, but it is mostly spent on infrastructure and operations down the line.

Since then, MindPal has evolved. It's now a no-code platform for business owners, entrepreneurs, marketers, creators, coaches, and consultants to build AI agents and multi-agent systems that automate their business processes.

We’re currently at $500k ARR.

An evolving product

Our early customers soon started asking for more. They didn't just want to chat with data; they wanted to customize how the AI behaved. Different users, like marketers versus researchers, needed the AI to respond differently based on their specific goals. We also landed our first B2B client, and they needed that kind of customization too.

That’s how we came up with the idea of custom AI agents in September 2023, and upgraded MindPal from a chat tool into a platform where users could build their own customized AI agents trained on their specific context and requirements.

While we were working on the product, we received a few requests for custom AI automation system development. One client, a healthcare company, needed to automate writing medical appointment notes, which involved several analysis steps. Another, a business coach, wanted to automatically evaluate business ideas using their specific coaching framework.

We realized these tasks were too complex for a single AI agent. They needed multiple specialized agents working together in a sequence. This led us to think of multi-agent workflows. And we decided to build that capability right into MindPal, allowing users to connect different AI specialist agents into a chain to automate much more sophisticated business processes. This was early February 2024.

Removing dependencies

In this space, AI libraries and frameworks themselves are often just as new and experimental as startups. When we started, libraries like Langchain on the backend and the Vercel AI SDK on the frontend were still evolving. This meant they weren't always stable, had hidden bugs, and frankly, there wasn't even a settled “right way” yet for how to structure things like AI agents.

We initially relied on these frameworks to get started quickly; however, we soon found that this reliance created its own problems, like performance bottlenecks and limitations on what we could build. For a platform like ours, where users need flexibility and control to build their own AI agents, this wasn't sustainable. We realized we needed deeper control over the core logic.

So, we gradually shifted, reducing our dependence on these external libraries and building more of our core infrastructure ourselves. Now all the AI agents and multi-agent workflows on MindPal are run in our own way, so we have the freedom necessary to keep innovating and iterating based on our own learnings without worrying about libraries’ limitations!

While we still use some tools like the Vercel AI SDK for some parts, owning the essential parts gives us the granular control needed to iterate fast, add unique features based on our own insights without worrying about external limitations, and ensure reliability.

However, the game is changing. With the rise of powerful AI coding assistants like Cursor, the purely technical roadblocks feel a lot less significant than they used to when we first started. Now, the bigger challenge — and what truly matters — is having the vision and the taste to understand what users actually need and how to build something genuinely valuable for them. 

Organic growth

Here's how we find customers:

1. Customer referrals: This has been a significant driver. We notice that when users successfully set up an AI agent/multi-agent workflow within MindPal that delivers tangible value and they use it frequently, they often start talking about it to their entrepreneur friends. This word-of-mouth marketing brings in highly relevant new users. To amplify this organic effect, we implemented an affiliate program with 20% recurring commission and built several sharing features across the app that give people more chances to share whenever suitable.

2. Community building (external and owned): We pursued a two-fold community strategy. Firstly, we partnered with relevant external communities where our target audience hung out to deliver presentations about how to leverage AI into businesses in general and how to use Mindpal in particular. Secondly, we established and nurtured our own community (the MindPal Hub Facebook group, now ~3k members). This space allows us not only to share updates and use cases but, more importantly, facilitates members sharing their own successful implementations and AI use cases with each other. This user-generated proof keeps MindPal top-of-mind for prospects within the group and continuously demonstrates the platform's value. 

3. Educational content (YouTube and blog): We invested in creating long-form educational content, primarily through YouTube videos and blog posts. The key here was focusing on practical, actionable advice centered around real-world AI use cases for businesses. We didn't just talk theory; we shared specific examples and often provided templates that users could directly duplicate and customize. This approach establishes MindPal as a knowledgeable resource, attracts users actively searching for AI solutions, and lowers the barrier to entry by giving them tangible starting points. 

4. Free tools (engineering as marketing): We developed and offered free, standalone mini-tools on our website, such as our free AI Agent Builder. These tools address specific, niche needs that potential customers might be searching for online. Users find and use the free tool, receive immediate value, and are then introduced to the broader capabilities of the main MindPal platform.

We’re also looking to experiment with other channels like paid ads and influencer marketing.

Revenue growth

Our business model is built on subscriptions. Customers pay us regularly to use the platform. We offer different plans — Individual, Team, and Business — based on usage factors like the amount of AI credits needed, how much knowledge data they store, and the number of user seats they require. Our pricing starts at $49 per month and goes up to $449 per month.

A huge part of our revenue growth strategy is making sure users get real value and stick around. We focus on a few things: 

1. Shortening time to value for new users: Since MindPal is a platform for building things, there's naturally a bit of a learning curve. We're constantly tracking and working to improve our onboarding so new users can build and run some AI agents (or multi-agent workflows) that actually work for them faster.

2. Making sure existing customers get value out of MindPal consistently and reliably: We pay close attention to platform performance — things like loading speed and AI reliability — and customers’ usage analytics. We need to make sure things that have already worked keep working and getting better and that all paid customers are actually using MindPal for healthy MRR growth.

About every six months or so, we also try increasing our prices, usually by 1.5x to 2x. We only do this after we've significantly improved the product and added significantly more value.

We were initially worried this might scare people off, but each time we've done it — moving from $12 (can you believe it?!) to $20 and $49+ now — we've actually attracted more customers, and importantly, customers who have carefully tested the app before subscribing, really appreciate the value and are a better fit for us long-term. Our early low prices just brought in folks who weren't serious and churned quickly. 

Good decisions

Here were our three biggest advantages:

1. My co-founder, Tuan: Honestly, the single best decision I've made was starting MindPal with Tuan. I really can't imagine a better co-founder. He has incredibly high standards of work. He’s extremely ambitious, so he's never easily satisfied – even when we hit a milestone, he's quick to point out all sorts of opportunities for improvement. I think it's fantastic because it reminds us there's always room for growth, like if we were this stupid and still achieved these things, imagine how much more we can achieve if we get better. His logical, first-principles thinking has been helping me see through surface-level ideas to what truly matters for our goals, or come up with better ways to do something. 

2. Using MindPal to grow MindPal itself: There was a point where we considered hiring more people to handle some mundane tasks, which can feel like the easy way out. But relying too much on hiring can make you lazy about figuring out if the work really matters or if there's a smarter way, maybe using AI. We started using MindPal intensively for our own operations – not just for simple things like writing blog posts, but for core, important processes where we could embed our knowledge and requirements into an AI agent. This forced us to think clearly about our own processes and document them properly. The results were often surprising: the consistency was undeniably better, and sometimes even the quality improved because the AI could handle vast amounts of information faster. More importantly, using MindPal ourselves helps us understand our product on a much deeper level, identify areas for improvement (serving ourselves better serves our customers), discover new use cases and possibilities to share with our user community, and ultimately make our customers stickier because they see more ways to get value.

3. Thinking in systems: It's easy to get caught up in the day-to-day grind and miss the bigger picture or smarter approaches, especially when we are a small team. Thinking in systems helps us discover smart ways to do things that provide exponential and sustainable results, much more than the conventional, default way we’re used to. For example, instead of just my team creating use case examples (which is slow), we built a system where our community can build templates, share them, and talk about them. This creates a passive, ever-growing source of new use cases exponentially faster than we could alone. This applies to so many more areas of the business than this. It forces us to think more deeply and effectively, not just react to the immediate moment.

Stay close to your customers

If I had to give one piece of advice based on our journey, it's to stay close to your customers. I think all reasons for failures of early-stage startups/businesses are some form of lacking customers.

At the start, focus on building something they genuinely need and are willing to pay for. The rest is just noise. You'll figure out other stuff down the line eventually anyway after you’ve got your customers. This is what I would tell myself if I could go back in time. 

What's next?

Honestly, I feel like we're really just scratching the surface of what MindPal can be, and there's still so much more work we want to do.

We strongly believe that a platform helping people launch their own teams of custom AI agents for their businesses is always needed as AI models get better and better every day. Because however better AI models get, they still don’t know anything about a specific business, its goals, its strategy, its processes, and human expertise and taste.

But we know we're not quite there yet on two big things we're actively working towards: 

• Making the building process 10x faster: Right now, it can take a few hours for someone to really wrap their head around MindPal and build something truly valuable for their specific business. Our goal is to slash that time down to minutes. We want it to be so easy and intuitive, that absolutely anybody can build their own army of AI agents with minimal learning and without needing much external help. And building custom AI agents becomes a no-brainer because it takes so little time for so much value.

• Making custom AI agents just work "out of the box": It's not enough to just let people customize agents. We want those agents to be genuinely effective and useful for your specific business almost immediately, understanding your context and needs without tons of tweaking. 

As AI gets better, it will handle more and more of the day-to-day execution in businesses, making things cheaper and faster. When that happens, what becomes even more valuable is the human element — the strategy, the creative ideas, the nuanced taste, the unique insights. We want MindPal to help accelerate that future. 

Basically, if AI can do something effectively, let it. Human time and brainpower are too precious to spend on tasks machines can handle. Some people worry about AI taking jobs, but we don't see it that way – we think that's missing the point. Just like with every other major technological shift in history, old or outdated tasks get replaced, freeing people up for new kinds of work, more ambitious projects, and bigger dreams.

Think about it – there are still so many massive challenges and opportunities for humanity, like truly exploring space, understanding the universe, curing diseases like cancer, figuring out quantum computing, and so much more. If we can automate the mundane tasks, imagine the freedom that gives us – the time, energy, and mental space to focus on those bigger, more important things. We're really excited to help enable and accelerate that shift.

Follow along

You can follow along on X, LinkedIn, and Github. Here's my cofounder's LinkedIn. And definitely check out MindPal.io and our YouTube!