Pricing an AI app without token anxiety

I had a conversation with a beta user who told me she'd stopped using a competitor's AI tool because she "couldn't stop thinking about the cost per message." She knew it was probably pennies. She couldn't help it. Every interaction felt like a meter running.

This is the token anxiety problem. And it's a product design failure before it's a pricing failure.

### The token anxiety problem

When users can see a direct connection between their usage and their bill — even at very small per-unit costs — they change their behavior. They ask fewer follow-up questions. They don't explore features that feel "expensive." They save up batches of questions to ask all at once, which produces worse interactions than the natural flow of a conversation.

Token anxiety is the opposite of what you want in a tool built around AI exploration. The value of an AI chat layer on your whiteboard summaries is asking the question that occurs to you right now, not the question you've been saving for when it feels "worth it."

A user who asks 20 good questions gets 20x more value from the product than a user who asks 2. But token anxiety drives users toward the 2-question version.

### Why flat pricing wins for most users

BoardSnap is $9.99/month or $69.99/year for Pro. Unlimited boards, unlimited projects, unlimited AI chat.

The cost is predictable. You know what BoardSnap costs before you open it. That predictability removes the meter-running feeling entirely. You can snap ten boards in a meeting and ask twenty follow-up questions and the price doesn't change.

Predictable pricing also makes it easier for teams to justify the cost. "It's $10/month" is a conversation. "It depends on how much AI you use" is not.

### The math behind the bet

Here's the honest version: flat pricing requires me to subsidize heavy users with light users. That's the nature of the model.

A user who snaps 30 boards a month and has 50 chat exchanges is more expensive for me to serve than a user who snaps 5 boards and asks 3 questions. But I charge them both $9.99.

I modeled this before launch using my beta data:

• Median user: about 8 board snaps per month, about 12 chat messages. API cost at current pricing: roughly $0.40–0.60/month.
• 90th percentile user: about 40 snaps, about 80 chat messages. API cost: roughly $2.50–3.50/month.
• 99th percentile user (power user): 100+ snaps, 200+ messages. API cost: $8–12/month.

At $9.99/month, I make money on the median user. I make less on the 90th percentile. I lose money on the 99th percentile.

But the 99th percentile user is the one most likely to recommend BoardSnap to their entire team. Losing $2/month on them might be fine if they convert three colleagues. The math isn't just "am I profitable per user" — it's "what does heavy usage signal about future value."

### The free tier limit

The free tier has limits for a reason: 1 project and 30 boards. These are calibrated so that real, ongoing use of BoardSnap as a professional tool requires Pro.

30 boards sounds like a lot. A team doing weekly standups plus bi-weekly retros uses about 6–8 boards a month. That's 4–5 months of runway on the free tier. Long enough to build the habit, not so long that you'd never need to upgrade.

The 1-project limit is the harder constraint. Real professional use almost always involves multiple clients, projects, or contexts. That's the trigger that converts free users most reliably.

### What I'd change

I think there's a version of BoardSnap pricing that adds a "pay-per-snap" option for infrequent users — people who have a quarterly planning session and want to snap 20 boards once every three months. $9.99/month doesn't make sense for them. $2 per session might.

I haven't built this because it introduces complexity — both in the payment flow and in the mental model. But it's on the consideration list.

For now: flat is right. Token anxiety is a product-killer, and predictable pricing is a feature.