Saagar was interviewed by Rory Woodbridge at "The ProductMarketer" to discuss all things Pricing & AI. Part 1 is about the fundamentals — how pricing works as a system, who should own it, and how to get started. Part 2 is about what AI is doing to pricing and the practical trends reshaping how companies charge for things.

What are the trends you're seeing across pricing projects? What's changing because of AI?

Seats-based pricing is becoming an existential issue for a lot of businesses. There are three key reasons why.

One is that with AI, we can now get closer to measuring value based on the work done, instead of just access to the tool. The second is the variable cost piece — how can you have a flat seat price when the variable cost means different users will use things at different intensities, and there's pressure on margins? And the third is more fundamental, which is that your AI products might — by their nature — be inherently trying to reduce the number of seats a customer has. So how can you tie your revenue model to a metric you're trying to reduce? You might be selling 10 seats today, but you're using AI to replace some of those seats. When you reduce from 10 to five, you make half the money while delivering double the value. That equation is existential.

But we don't say it's one-size-fits-all. Seats aren't wrong for everyone. Take Slack — they still use seats. Their AI isn't trying to reduce the number of people you need. It's a collaboration tool where everyone's on there, and more people being on there makes everyone a bit more productive. That's what we call more of a co-pilot type of AI, and the unit of value is still a human or a seat. The difference is you've still got to monetise the premium version with AI versus the base version without it.

People are talking about outcome-based pricing, especially with AI products. Is that realistic?

Some folks see the world moving towards outcome-based pricing, where you're creating a P&L outcome for the customer — cost savings, revenue generated, whatever it is — and your AI is helping to get there. That's a good aspiration. But, at Northlane, we think the world is a long way from that as a default pricing strategy.

The reason is mainly attribution. You've got a BDR tool that you think helps close five more deals per month. You can make a business case that your product has driven 100K of value in a year. But for you to then say 20% of that is coming back to you — a lot of things must go right for that attribution to hold. And what CRO is going to give that money back? They're at the board meeting getting congratulated, and then you're saying 20% of this growth was down to the tool. That's a tough political conversation.

There are a few situations where it can work, though. One is where you're connected to the payment loop. Companies like Chargeflow, for example, who go out and collect payments that have bounced. Their agents do that, and they come back and say they've found £10,000 you were never going to get. So, they keep £2,000 of that — because ultimately you would have got zero without them. When you're close to the payment loop, it makes a lot more sense. It's a narrow win-no-fee setup.

So usage and consumption pricing is the middle ground. How sustainable is that?

Credits-based models have always been part of pricing. Think of Audible, or old mobile phone contracts — you'd top up 10 pounds and draw it down based on a minute of phone time or a couple of text messages. It's not new.

The difference is that credits are now a way to communicate something with a variable cost that might also have variable value. Each unit can be assigned a specific value. But the challenge is that the variable cost basis is continually changing. If Anthropic or whoever changes their pricing structure, you must keep rebasing your credit system when different actions cost different things and you're trying to preserve a certain margin.

What I see as more of a risk is that 'cost-plus' thinking is back in vogue. We spent a lot of the SaaS era talking about value-based pricing, and token-plus pricing gets us back to cost-plus. For an early-stage business, you've got to stay alive and cover your costs. But as soon as you've got product-market fit and you're clearer on your value proposition, you've got to orient the model to communicate value. And credits can obscure that — you cloud the specific value props within the credits system.

Clay is a good example. One of the key reasons they changed their pricing recently was to elevate the value proposition of the orchestration layer they're providing. Before, it felt like credits connecting you to various systems, and they came across as a data infrastructure player. When really, where the value sits for them is in the connective tissue across all those systems. Being able to distinguish value from cost in your model is going to become even more important.

It sounds like we're watching the SaaS story — working out pricing from scratch — play out again, just faster.

Exactly. Moore's law would have us think that compute costs and token prices are going to trend down. But that's only true once we've exhausted the quality constraint. In the SaaS and cloud world, when it was as good as it was going to get, costs started trending to zero. But we're still in a world where everyone is trying to make better models, make them run faster. There's so much efficiency growth at the language model level. So, we're going to see significant fluctuations in price, and because the timelines are so much quicker, the fluctuation gets compressed.

The best trend we're seeing is people treating pricing like a product. You iterate, you get feedback from the market, your customers, your finance team. A lot of companies were sleeping on their pricing model for a while, maybe doing price increases occasionally, but not thinking about it as a design system with loads of different inputs. Treat pricing like a product and make sure you're still engaged with customer value.

The data shows companies made 3+ pricing changes on average in 2025. Is that a good trend?

We're seeing that. Some indices show companies are changing prices about three times a year now. And that might be price points, which are more nimble changes. But it's also packaging structures and price models, which are more fundamental shifts.

That shows what we've been saying — treat pricing like a product — is now happening. If you're getting feedback from the market on your product, you'd be more than happy to ship a new version when the time felt right. But you've got to have the dynamism both upstream and downstream to make that work. There's the tooling side — billing systems, data, telemetry. There's a whole new ecosystem of tools recognising you need to be nimble. And upstream, you need to set expectations with the execs that this is not set-and-forget. Set up a governance and pricing committee. Bring data every month. Here's what we're getting from the market, what the sales team thinks, what customer research shows. And then you say, all right, stick or twist. What are we changing?

Companies like Clay, Notion, and Figma are experimenting with their pricing as a point of pride. It's almost their strategy. I believe a lot of product marketing thought went into how Clay changed their pricing, because it was all about positioning. They were being viewed as a data layer, when they said our value is in this orchestration layer. And that's very much linked to positioning. The most impressive companies have that constant iteration loop between who they want to be and how pricing evolves with it.

When pricing changes, there's usually a mountain of product marketing work. What do the companies that handle rollouts well do differently?

Getting buy-in and not having it be a big surprise. Some folks have this idea that they're going to build a new pricing model in stealth and launch it, because they can't be bothered taking everyone on the journey. I don't think that's the right way at all. It's about bringing people along, getting trust and credibility, and then having the speed of iteration to take feedback quickly. Recognising what's good feedback and what's noise is a muscle you build over time.

We'd also recommend testing with new customers first if you have a new business engine firing. That's your quickest way to get feedback without being scarred by legacy packages. You can assess whether the new pricing model fits the new world. Then it's a separate question of how to migrate the old world onto the new one.

Price transparency feels like it's becoming more important, especially with AI-powered search. Are you seeing that?

This is becoming a hot topic again. With more answer engine optimisation and agents doing procurement work, if they can't find your pricing on the website, you're going to get kicked out of a lot of conversations. That's going to be table stakes for go-to-market — having a clean, clear pricing page with strong messaging. And maybe even price points on the website, which is a perennial debate in the pricing world.

If you see that world moving in the direction of agents talking to answer engines, then the shopping cart after the comparison is probably an easy next step for the big tech companies. You see a comparison of three products in an AI overview, and you could easily see a buy button. That's already happening in the consumer world, and there's no reason it can't happen in B2B software. So, price transparency is going to become increasingly important.

For a product marketer who doesn't have budget for a specialist — what can they do to get smarter about pricing?

There are some strong resources out there. Kyle Poyar's Growth Unhinged is excellent. The Pricing SaaS guys, Rob and Jon, are doing good work. And Lenny's Newsletter often has people talking about pricing as well. Madhavan Ramanujam's Monetising Innovation is worth reading — it breaks pricing down to first principles with strong analogies and examples.

But what I would caution is that theory is one thing and practice is another. Be wary of seeing hot takes on LinkedIn and applying them straight into your business. There's so much more upstream and downstream work needed to make a piece of theory apply to your specific situation. And that's where external help does add value — someone who can see the bigger picture and come at it with a fresh perspective built on pattern recognition. No one has done your specific job at your specific company at this specific moment. So don't beat yourself up if the framework you're hearing doesn't quite make sense for your situation.