Madhavan Ramanujam: prediction

Transcript context

…So when you talk about AI companies and monetization models, we get asked this question. "Should I be usage-based? Should I be outcome? Should I be a copilot mode? Or how do I actually think about my pricing model?" So we came up with this framework, which is a relatively simple, straightforward framework, but very powerful. So there are two axes here. One is attribution, and the other one is autonomy. And when you have high attribution and high autonomy, that is when you have high pricing power, and we'll come back to that in a bit, right? So let's take the first bottom-left quadrant. That is the quadrant where your attribution is low and your autonomy is low. In that situation, the best pricing archetype that actually fits is it is actually a seed-based or a subscription model because there's not much to do about it because you're not being able to attribute a lot of value to what you bring, but you're in a copilot mode and you're not in an autonomous mode, so a seed-based pricing would actually make sense. But if you're at that quadrant, the immediate thing to think about is, how do you actually build more attribution and move to the right so that you actually get more pricing power? So if you think about the bottom-right quadrant, those are companies that have actually done that. They can prove more attribution to what they actually bring to the table, but they are still not in a fully autonomous mode, as in there is still humans in the loop. If you take Cursor, for instance, it definitely improves productivity, can actually bring down the time to actually do code, the attribution is clear, but it's still in a copilot mode. In those kind of situations, a hybrid pricing model is the best option where you still have a seed-based model for the copilot kind of use case, but you also layer in a consumption model which actually says there are a certain number of AI credits or tokens that can layer in the usage aspects. So if you use more and more, then you're actually paying more on consumption. So it's a hybrid model that actually works there. If you look at the top-left quadrant, those are products that are very autonomous but are not strong on attribution. So these tend to be mostly backend or infrastructure kind of products that are core critical to run businesses, can be autonomous, but they're not directly impacting the KPIs that businesses are tracking, and hence cannot prove attribution very effectively. So in that situation, you need to be on a pay-for-what-you-consume and a usage-based model. A seed-based model would not make sense because it's autonomous. There's no human in the loop. But that's why you're also in a usage-based model, saying the more you use it, the more you're actually charged. And usage becomes a proxy for the value that you bring to the table. The quadrant that you really want to be in is the golden quadrant, which is the top-right one. t, the more you're actually charged. And usage becomes a proxy for the value that you bring to the table. The quadrant that you really want to be in is the golden quadrant, which is the top-right one. That's the outcome-based pricing model where you have great autonomy and great attribution. And here is where I think AI can be really magical. So this means you're not only charging for work delivered, but you're charging for work delivered that was delivered by AI without no humans in the loop. So that becomes more of an outcome-based model situation. So a classic example here is Intercom for Fin. What they actually do is they charge based on an AI resolution. So if an AI is able to resolve the ticket completely independently without a human in the loop, then they charge for it. If a human intervention is needed, they don't charge for it. So they're more on an outcome-based model. Or companies like Charge Flow would charge up to 25% on a charge bag that they're able to actually recover, because these are core savings that you actually bring to the table based on your AI. It is highly attributable, highly autonomous so you can start moving towards an outcome-based pricing model. If you look at the state of where AI is today, as of the day of recording this podcast, now, the most popular model is right now a hybrid pricing model. So because this is also expected, because the previous SaaS playbook was usually on the seat-based model, but they've all now moved on to at least a hybrid to actually incorporate AI credits and usage, et cetera, about 5% of companies are probably in a true outcome-based pricing model as of today. But those companies, some of the best ones are able to recover 25 to 50% of the value that they actually bring to the table. In the classic SaaS situation, we used to say if you can charge 10 to 20% of the value, that's actually great. But in AI, you can actually charge 25 to 50% because it is autonomous, you're doing it with the AI. There's no humans in the loop. You are creating incremental value to the business metrics. You are producing hard-cost savings. There's opportunity costs. You can justify all of that. It's attributable. So you can actually take 25 to 50% of what you bring to the table. In a lot of benchmarks and studies that actually show, and this is also my belief, that in the next three years that 5% number will move to 25%. So what this really means is if you want to win in AI, figure out a way to get to that quadrant because that's a magic quadrant. If you can truly price based on outcomes, you've achieved and unlock tremendous value. Wow. Okay. I'm just going to pause again. This is amazing. Madhavan, thank you so much for it. I love, again, that you just spent years, decades studying this stuff, come here, tell us all the answers of what we should be doing. This is incredible. Let me ask you this. By the way, for folks not watching on YouTube, the companies you have in the golden quadrant, outcome-based pricing, Sierra, Fin, and Chargeflow. We've got the founder of Sierra and Fin coming on the podcast soon, so we'll talk about all this with those guys.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.