Pricing AI agents: What does ‘value-based pricing’ really mean for AI?

As builders, we obsess over the AI experience – how smart it is, how fast it replies, how many conversations it resolves. But none of that matters if we don’t get the pricing right. If it feels unfair, unpredictable, or misaligned with value, adoption stalls.

To guide our pricing strategy, we’ve always been strongly led by customer feedback. Intercom had the first AI agent to pioneer value-based pricing, a decision that we came to after extensive research.

We just completed more deep research with decision-makers buying AI agents. Here’s what we learned.

Most buyers don’t want to pay based on usage

Let’s start with the obvious one. Many AI agent providers charge per conversation. It’s clean, familiar, and sounds simple. But not a single buyer we spoke to actually wanted to pay that way.

It’s not that conversation-based pricing is inherently bad – it’s just not value-aligned. Buyers rightly pointed out that with this model, you pay whether your AI is useful or not. One buyer put it like this:

“It’s like calling a call centre and the agent says, ‘you suck.’ That counts as a conversation, but it sure as hell shouldn’t count as value.”

So why is conversation pricing so popular in the market? Because it mitigates risk. It helps companies get started with AI when they have no historical data, no sense of what “normal” usage looks like, and need to get something live. When there are lots of unknowns, it feels like the most secure option.

But is it really the best way to manage uncertainty – or just the most familiar?

There are better ways to solve for this uncertainty without designing your pricing metric around it. Giving customers the ability to test your product and the flexibility to scale usage without penalties is a far more customer-obsessed, value-aligned way of charging for AI agents – while still managing risk.

Outcomes are the clearest path to value

For Fin AI Agent, that’s resolutions.

The most widely-preferred pricing metric in our research was resolutions. It just makes sense: you pay when the AI solves a customer’s problem. If it doesn’t, you don’t. That’s aligned. That’s fair.

Resolutions also create healthy pressure on the vendor (us) to improve the product. And they give buyers a clean way to compare ROI: how much does it cost to resolve something via AI vs. via a human?

But it’s not perfect. Predictability is still a concern. Because resolutions is a newer metric, many buyers don’t yet have enough historical data to forecast their bills with confidence. That’s something we, as vendors, need to help solve.

Trust is the currency of AI

We also tested a more advanced concept: AI-confirmed resolutions. It’s like resolutions, but with a second AI reviewing the conversation to determine whether the issue was actually resolved.

Some loved it. Others didn’t trust it.

On the surface, it gets us closer to true outcome-based pricing. But as one buyer said:

“How do I know the second AI isn’t just agreeing with the first?”

There’s a trust gap here, and we shouldn’t ignore it. We’re asking people to believe in not one, but two black boxes.

There’s potential in this idea, but when pricing new products, we have to start with metrics that feel fair and earn trust.

Total price still trumps all

Here’s the reality check: while buyers care about fairness, predictability, and alignment – they care about cost savings more.

The main reason people buy AI agents is to reduce support costs. When the rubber hits the road, they’ll usually choose the cheaper pricing model, even if it’s less appealing.

We saw this again and again. Buyers said they preferred resolutions – but if it was 20% more expensive than Conversations for the same product? They’d go with Conversations. A few said they might pay a ~10% premium for the “right” metric. Beyond that, price wins.

This has huge implications for vendors. Even the most elegant, value-based metric won’t land if it comes with an uncompetitive price point.

So, where does that leave us?

We’re still early in this journey. Pricing AI agents is complex, and there’s no one-size-fits-all solution. But here’s what we’re taking forward:

✅ Resolutions is the right direction – but buyers need help getting there, support, data, and transparency.

💰 Price still wins. A great metric won’t land if the price point doesn’t work.

🤝 Trust is the currency of AI. Buyers need to trust the agent and the pricing model.

Value-aligned pricing isn’t just the future – it’s the benchmark.

This isn’t just about pricing. It’s about trust, alignment, and long-term partnership. And we’re closing the gap.