Monetizing AI — Alvaro Morales, Orb

[on-hold electronic music] Hi, everyone.

My name is Alvaro, and I'm excited to chat today with you all about pricing strategies and how to think about mon-monetization, uh, specifically around AI. This conference has and continues to be amazing where I'm walking down the hall and seeing innovation everywhere.

And th-th-an idea that I think is really important is for there to be a successful innovation ecosystem, we also have, have to think about how we capture that value and monetize effectively to reach the right group of folks and make sure that this becomes a s-self-sustaining and, uh, effective ecosystem.

I am the co-founder and CEO at Orb, and we specialize in billing for AI and SaaS companies, and we've been immensely lucky to partner with companies all across the AI stack, and we've been, uh, a small part in helping some of these amazing companies ship incredible product experiences like Vercel's v0, Replit Agent, and Perplex-Perplexity's API.

So with this experience, the goal of my talk today is to share some stories from the field and bring forward, uh, some of this bleeding-edge, uh, experimentation and ideas that are forming on the world of AI pricing.

And I think a good place to start is to really think about what is unique and complex about AI pricing specifically. I think software pricing has always been a little bit of an art and a science, and there's always been some considerations around how to get it right.

But there are three things in particular that make considerations around AI pricing perhaps more challenging than they've been before. Well, first of all, things are changing dramatically quickly in terms of model, model and inference costs, in terms of capabilities, in terms of new products being launched.

So, uh, just the, the pace of this evolution makes it hard to keep pace on the pricing side.

Second, uh, these AI technologies create some real pressures on margins and COGS. So the thing that we could do maybe ten years ago of, like, shipping our products and we'll figure out how to price them down the line is much harder to do when that Anthropic or Open A- O-OpenAI bill is very steep and significant.

And then third, amidst all this experimentation exploration, I think customers are really, really hungry to understand what is the ROI that they are getting from, from some of these technologies.

So this makes AI pricing uniquely challenging and, um, this is even greater so when we keep getting blown away and surprised by the kind of, um, adoption trends and just uptick that a lot of these products have.

When I say that AI pricing is, is challenging, I really mean that it's kind of challenging for everybody and where, um, notably we've seen even cases where, uh, ChatGPT Pro at two hundred dollars turns out-- turned out to be a, a loss driver for the company.

So I come today to you with an idea that perhaps as an industry we should not be pricing on vibes and there is a potentially better way to do this.

So to drive my point forward, I have come to you with three simple frameworks and one tool to share with you some ideas and, and, and frameworks for how to think about AI pricing and, uh, share with you a tool we've built at Orb that is really helping some teams make the right decisions around their monetization.

So first and foremost, like, should you monetize AI? Um, and how should you think about, uh, that, uh, that experience? So over here, um, I have a kinda simple framework, uh, that is one of my favorites.

It's, it was created by the folks at Simon-Kucher, and it really starts asking about this idea of how, um, should you think about monetization. Should it be a direct monetization strategy where you are either selling something as a standalone product or charging it for it as an add-on?

Or should it be an indirect monetization strategy where perhaps you're using it to drive upsells into existing products or tiers that you have? Or maybe even bundling it for free because it's incentivizing the right behavior you wanna see in your customers.

So let's look at some examples. This is G-GitHub Copilot. Uh, they launched it as a separable monetizable add-on on top of the base GitHub seat. An important consideration here is, is this, this new product, is this new feature adding value to everybody in your audience or just a niche group of folks?

Um, and when it's maybe not everybody with a GitHub seat is gonna get value out of this, maybe an add-on monetization strategy is a really good one.

Second, second example here is Notion, uh, and Notion AI. Notably, they were first to market in their category and initially launched Notion AI as an add-on like GitHub Copilot still is.

But recently, they've bundled this into their business and enterprise tiers to kind of encourage, um, more adoption there. And as part of that, uh, they've done some price increases on those tiers.

So they're, they're finding ways to kind of encourage the, the adoption that they wanna see at those higher tiers.

Finally, this, this is a, a little bit of a more off-the-beaten-t-path example. A month ago, Expedia launched a new AI-powered feature that lets you turn Instagram Reels into bookable trips, and they're not charging for this.

They have, like, released this out for free. They're doing this while swallowing those, like, cost to serve and model inference costs to power this capability because likely they are hoping to see a lot more bookings of travel as a result of this capability.

So again, this is an example of cases where you might even wanna indirectly monetize, where you're not charging for this right away, but you are incenting the behavior that then will continue to fuel your revenue growth.

Having explored, like, should you even monetize this, how do we actually do it? And I think one of the most important considerations here is to make sure that you are picking and selecting the correct value metric.

So let's take the space of AI agents, where, uh, literally AI agents are redefining the way that we get value out of software. Where before we measured the value of software by getting a login into a web application that we could, like, spend our days in, now we have AI agents that are out there doing work for

us or achieving automation or achieving outcomes for us. How do we price that? Um, I think what we're seeing out in the field, es- especially over the last six months, is that there's a little bit of a spectrum in terms of how tightly aligned to discrete units of value or more closely aligned to ROI do you want

to be. And there's sort of a spectrum between very resource-based or token-based monetization models, proxies of value, like steps in a workflow, maybe entire workflows, perhaps e- even more direct labor replacement type strategies where just like you would hire a consultant per hour, maybe you would hire an agent per hour, all the way to true outcome-based pricing.

Let's run through some of these examples again. So here's Vercel's v0. They have a token-based model, uh, based on very granular capabilities, and I think this aligns really closely with their audience, the developer that I think really wants to understand some of the capabilities that they can drive over there.

On the other hand, here's Zapier. They, uh, deliver automation at scale, and they've gone with a task-based pricing strategy where various tasks in, in an entire workflow or Zap are the unit of value that you end up paying for, and you actually subscribe to a certain number of tasks per month or can go over if you are

exceeding that. And then finally, there is true outcome-based pricing. So this is Intercom's Fin, where the way Fin is priced is it charges 99 cents per successful customer support ticket resolution.

Where literally if the end user at the end of that support interaction says, "Yes, that answered my query, and I don't need to escalate to a human support agent," that's where Fin charges for 99 cents.

And if you've been on LinkedIn, you've probably seen a thing or two about outcome-based pricing because I think it is very much actively being discussed and actively being explored.

Maybe the hidden secret is that outside of this customer experience category, customer support space, it's really hard to find, um, many examples of this working really well quite yet.

So perhaps this is still kind of more on the emerging side. And why is that? Because ultimately, customer and vendor need to align on what the definition of the outcome is and need to, uh, furthermore be able to measure it in a somewhat objective way.

So when you're dealing with a customer support ticket interaction that's very clear and simple, you can ask, "Hey, did this resolve your question?" And you can see whether this needed to be escalated.

But whenever you're trying to proxy things that are maybe more, um, creative or perhaps more, um, you know, less directly tied to, uh, dollars and cents as, as can be measured by the cost that support teams incur for businesses, that's, that's been a little harder.

So ultimately, what I find very inspiring and exciting about outcome-based pricing is that it's not a proxy for R- ROI, it's literally ROI. So I think as an industry this year and, and, and for many years thereafter, I think we're going to continue trying to find ways to crack this code and find, um, the right way to

align outcomes and measure outcomes such that, um, this aligns with everybody.

So lastly, having explored like a couple of frameworks and ideas for how to price, uh, I think what's really important is to build a muscle around thinking about pricing strategy as something that needs continual evolution and experimentation.

Maybe 10 years ago when we were building SaaS, what we could do is like ship a product, pick or guess a price point, and then spend, you know, a year, two years, three years building and adding value to that product.

And then we might realize like, "Oh, you know, you know what? We actually have to catch up to all this value that we've created and, and think about doing a big, large, painful, um, price increase."

But nowadays with AI, when, um, you know, new model drops can reduce cost by 10X or increase cost by 10X literally overnight like Claude 4 did, I think this, uh, kind of one, two, three-year cycle of pricing iteration just doesn't cut it anymore.

So what we see is that the best teams out there are able to find ways to, um, really drive a lot of experiment- experimentation and continual evolution in their pricing.

And as we've been talking about pricing strategy, I think it's important to remember that it all comes down to not just the price point, but rather the cohesive entire experience around pricing.

Everything from how you package features into different tiers or how you think about rate limits or volumes or even custom terms. How you design this, but more importantly, how you communicate this to customers contributes to that overall alignment of price to value and should be something that, uh, that we find teams need to and should be c-

continually experimenting around. So how is this actually done, and, and, and how are we finding, um, how are we helping teams do this? I wanted to share a, a little demo with you of Orb Simulations, which is a product that we've built in, a- around this idea of helping, um, customers find the right AI pricing strategies.

So let me tell you a little bit s- of a story here. For context, this is Orb. We are connected and integrated with a raw and real-time stream of product usage events where what you're measuring and instrumenting in your application, you can send into Orb for billing.

And, you know, we've created a very flexible billing offering where, um, where, like, folks can come, come in and, um,

be able to, like, monitor in real time what their accrued costs are and get a sense of, um, how much charges are customers racking up. So again, this, this idea is like the marriage of an analytics product with a billing product connected together to really drive a lot of that real-time insights.

But last, uh, last year in, in about the last six months, what we were noticing was a lot of our customers were using our product in a little bit of an unexpected way.

What they were doing is they were sending data into Orb, and then they were setting up shell pricing structures on top of that data, but not actually billing their customers for it.

And we didn't understand why they were doing that, so we called a few of them up, and turns out that they were running the closed beta periods of a few upcoming product launches, and several of those were the, the closed beta periods of, uh, AI agent products coming out to market.

So these teams had a product build. They had a group of customers that they, they wanted to start to get feedback from but weren't fully pencils down on what exactly was the right pricing strategy, and they were trying to hack Orb into figuring that out.

So what we did is we built a whole product experience around helping that, that workflow. So what Simulations let you, lets you do is back test or, or, or test out alternate pricing strategies on top of that data platform of, of product usage, and it's designed to help teams really answer that what if question.

What if we tried pricing that way, or what if we experimented with a different model? Let me show you how that works. So in Orb, you can come in and define a simulation.

So maybe for our, our demo over here, we're gonna pick a simulation time period that perhaps is gonna span the period of our closed beta, and we're gonna pick a cohort of customers.

This might be just customers that are not yet paying in that closed beta period or might even be a group of existing customers that are, um, paying for a previous or older version of, of your pricing and that they're about to get, uh, a new kind of value.

And what we enable you to do is to build out scenarios for alternate ways to price. So maybe we can start with our baseline, and what we're gonna do is try more of an add-on based pricing strategy, where what we'll do is, um, uh, maybe come in and say we wanna add a fixed fee.

We're gonna charge everybody $20 for, um, AI agent access. Alternatively, we might wanna add a more, uh, granular or discrete pricing strategy where we can come in and actually charge for tokens.

And perhaps for those tokens, we want to charge a simple unit price or maybe more of a tiered fee, where some tokens come included for free, um, and others are charged on top of that.

Again, the idea here is to really help you build a different set of pricing possibilities so that you can help answer your ideas about how to price in a rich data in- informed way.

So when we run that simulation, what we'll generate is a report that shows you, uh, a lot of insights and, and da- and, and angle points into this decision of how you should price.

So first and foremost, between th- these scenarios that we've out- outlined, we'll show you what the biggest impact to top line revenue is, so you can kind of understand based on the actual consumption usage patterns of your customers, uh, which, uh, pricing scenario might, uh, be the most effective one.

But if anybody who's lived in this world kn- uh, knows and understands that it's not just about top line revenue. When you're dealing with existing customers that are paying you a certain way or a certain amount, it's kind of hard just to hike up pricing on them.

So getting an understanding of what the lowest average change to existing customers will be is also really important. And so what we give you is a very detailed view into the revenue mix that you might expect based on a particular pricing strategy and even a scatter plot that shows you, like, percentage magnitude change as well as revenue

impact for different customers, uh, obviously with, with an exportable data set of impact to each individual customers. I think the big idea here is to break through a lot of this fear, uncertainty, and doubt of how should I price with data and help teams, uh, you know, if you will, travel time or, or really help explore the

space of pricing possibilities so that when you get to that moment where you're ready to launch, you're doing so very confidently and with certainty of what that impact is going to be to your existing customer base.

So that's Orb Simulations. And again, really the idea here is if we really want, want to help you not have to price on vibes or, uh, uh, ideally, we'll give you some tools and our perspective is that, um, uh, AI builders should always simulate first before putting something out to market.

To close up over here, you know, we've explored some, uh, ideas around how to price. First of all, thinking about should you even monetize directly or indirectly. Then explored the value metric selection question, just making sure that the way you're pricing really aligns to the value that you want your customers to perceive from your product and to

incent the behavior that you want them to, to take. And then thirdly, um, really explored how, uh, experimentation and thinking about pricing evolution in a continual way is not just valuable, but really critical and important, um, for, for these kind of AI products.

So you might be thinking that this is, um, hard or tricky. Well, it is hard or tricky, but, uh, what we've seen with our customers is that they've been able to achieve truly headline inspiring results on their revenue front and get their amazing products to the right audiences by finding that, uh, aligned and, and correct pricing strategy.

Um, and we've really been able to see up close that when you can get it right or when you can converge to right, the impact can be huge and, uh, valuable.

We are gonna be hanging out today on booth G17, so if you're curious about how you should bring your AI... your new AI products out to market, or if you want more data points or ideas about how folks are thinking about AI pricing, um, we'd love to chat.

Thank you very much. [outro music]