Adam Thier 0:00:05.3: 

Good morning, Boston. How are you guys? So if you can tell, I'm not from Boston. Can anybody guess where I'm from? No. My grandfather was from Ireland. Texas? Yes, that is really… No. I was born in Albany, halfway between New York and Boston, so I'm New York sports team, and obviously a Boston accent. So first, I'd like to bring attention to the latest in Anaplan collectibles. The llama. If anyone can tell me why we picked the llama, I would like to know. We've been searching on AI for correlations between Anaplan and llamas. They're evidently very obedient - which doesn't apply to me at all - trustworthy, they can carry a heavy load, but I'm very excited. So I'm going to keep my llama right here, and when you go to lunch, you're all going to get a llama. Oh, I'm in charge of clicking today. All right. I'm here to talk about what? Have you heard about this AI stuff? I mean that joke is getting so bad at this point. Has anybody heard of AI? How many people in the room are actively using AI? Yes. How many people in the room are getting like, they feel material benefit out of AI? It's a little bit less. How many are getting a lot of pressure to do more with AI? Yes. Me too, by the way.  

 

Adam Thier 0:01:46.4: 

So I've been doing planning - if you've seen me up here before, you know this - I've been in the planning space for 26 years. So that's a long time. I wrote my first planning system while crossing the Atlantic with Christopher Columbus to discover Americas. So 26 years ago, Michael Gould and I, and Peter [?Warri 0:02:15.7] and John [?Sandos 0:02:17.3] were sitting down and building the very first, what became Anaplan. It was Adaytum. Then I went to Hyperion, and I went to SAP. So if you're using BPC or Hyperion, that's got my paper cuts all over it as well. So I've been doing this a long time, and I have to tell you I haven't been this excited in a long time, because AI happens to fit what we do well. You can view AI in a whole lot of ways, right? Is it like a bunch of extra people, Silicon people, they're going off and doing tasks for me? Yes, it is. But ultimately what we do in planning is we help you make decisions. What are we going to do next? And that's why we call it decision excellence, and what is behind an excellent decision? I mean, I'm the king of bad decisions but hopefully with AI, I'll make fewer of those. The key to great decisions is having all the facts, all the data. So a lot of things are intersecting at Anaplan right now to collect all the facts, put them in something where it can analyze all the facts and run a number scenarios and models for you and with you to come up with the best possible scenario, because remember, we're not trying to make the best decision, we're trying to make the least worst one.  

 

Adam Thier 0:03:40.0: 

There are always lots of good decisions. You try and figure out which one is the least worst decision for your business, right, because we want to build these products while it has working capital implications, we want to venture in these territories. Well, what do we have to do from a workforce standpoint? What is that going to cost? Well, you look at Anaplan and you look at the applications, and you look at the underlying platform, it's all around getting as many facts as you can, and analyze them really quickly, and what we've been doing to do this is we've been spending $500 million. So somebody, and I don't know what they were thinking, gave me $500 million to invest in innovation. We invested in three things. The platform. So the Anaplan you know and love today, be it the Classic engine, which we still continue to invest in heavily, the Polaris engine to scale bigger, ADO for getting all that data. I mean we have customers - we have one customer in Europe that's using ADO to bring data in from over 200 systems, 100 million rows a day that's got to be brought together, synchronized, harmonized, and fed into Anaplan. That's the spectrum of the problem, in order to get all of those facts so that we can then apply AI to it, and between the platform and AI are the applications.  

 

Adam Thier 0:05:05.3: 

The applications are there to reduce your cost of getting the Anaplan and get more Anaplan, better, because, again, a great decision requires great data from everywhere. So if it's not in Anaplan, it's not something you can get in Anaplan, that fact becomes less usable. So everything we've done, the $500 - we haven't spent $500 million yet. That doesn't conclude until next year, and just gives you some context, over that same period of time, the $500 million isn't everything we've spent on R&D. We will have spent over a billion dollars on R&D. So half of it is going to innovation. That should give you the context of how big Anaplan has become, and we've got that big because the people in this room made us that big. We didn't do it ourselves. You spent the money with us, because obviously we were adding value, and that next generation of what we're doing with AI is all about rapidly bringing that data into Anaplan from everywhere and being able to run those scenarios and adopt it. So I'm going to do a live demo in a few minutes. So I'm going to fly without a net here, and I'm going to show you our latest AI innovation [?launch 0:06:18.7]. So are we good? $500 million. We understand the problem we're trying to solve. It's this big trying to get it to this big and just helping you make better decisions.  

 

Adam Thier 0:06:30.5: 

So let's put some context around AI, because we all knew and heard about AI, and then two and half years ago the chat engine showed up, and all of a sudden chat equals AI. But that isn't the only part of AI. So when you think about Anaplan, we've been in AI for a long time. Michael Gould was a mathematician. I'm a Bayesian, if anyone knows what that means. But we had PlanIQ forecast three years ago. We've had Optimizer, right? We had the APIs. Lots of people have been using data sciences alongside Anaplan for a long time. But that was kind of Phase I, right? You know that AI ML forecasting and optimization, allowing those numbers to get crunched faster. Last year we rolled out our conversational interfaces, the CoPlanners, the two little red dots in the corner where you can ask it questions, and it's trained on over 800 questions. It can literally answer hundreds of thousands of iterations of, what's my most profitable product? What's my forecast for that most profitable product? What is my inventory of those most profitable products? Have that entire conversation. How many people do I have trained to sell or implement those most profitable products? All to be able to have that conversation with Anaplan. So we're no longer spending a lot of time building static reports and dashboards to do it. We can literally answer those questions in real time and then take those questions - and we do this within CoPlanner - and add it back to your dashboard so you know what's going on.  

 

Adam Thier 0:08:08.6: 

Phase III, which we're in right now, is the analysts. So somebody that sits, is a true agent that acts on your behalf, that goes and gets things, that answers those questions and goes out and finds more things for you. We have one of those for finance, one of those for sales, one for supply chain and workforce, and then we're going to get to, in a moment, CoModeler. Phase IV is when we really start to accelerate, and that happens at the end of this year, anomaly detection throughout the platform. So all data coming in through ADO will be screened for historical precedent, and if it's outside certain standard deviations it's going to flag you as anomalies, and then it's going to either - is this a data quality problem, or is this a new trend in our business? So a lot of those things. The profitability monitor, workforce status, sales quota management. All these things move from reactive to proactive with AI. Then finally, coming out next year is our Agent Studio, allowing you to build your own agents. Make sense? 

 

Adam Thier 0:09:14.5: 

So one of the reasons Anaplan is able to do this is because Anaplan has always been AI at the core. What does that mean? We're infused with AI or…? No. We're AI at the core. So if you look in the bottom left corner, Polaris Linear Algebra engine, and, you know, we didn't have room, but in the Classic there's also a linear algebra engine. So if you know AI, you think about a chat engine. Well, a chat engine is just a module that's been trained on a neural net. So these guys create these huge neural nets that go out and scrub the internet, and come back, and they count a lot of things, and they create probabilities and statistics. Then when you talk to chat, it says the next word is probably this, or this is what they want, and just create a huge interface. But that AI module is based on a neural net, and what are neural nets built with? Linear algebra.  

 

Adam Thier 0:10:12.4: 

So Anaplan has always been an underlying linear algebra, and you can see it when you think about Anaplan. Inside the Anaplan calculation are a bunch of - that's why we use honeycombs - a bunch of beehive frames. So you have products and time. So that's a frame of data. Then there'll be people and salaries, and you have all those frames, and what happens is they all exist together in that engine, and they intersect, and they cross each other, and Anaplan's special sauce, which nobody else has been able to recreate is the ability to handle quintillions of cells in those little honeycombs, map them so you can access them really quickly, and recalculate them. Again, you saw the data before, but we have one customer that's got 600,000 sales outlets and 10,000 SKUs. Every hotel bar, every bodega, every liquor store, every restaurant, every place that sells alcohol across 19 states, and they can see in real time what their market share is versus everyone else, and then when they want to do a promotion, they can do it specifically at the SKU level at the location level, and they do that every week, and when they go into model simulate that, it's a couple of seconds. Each time everybody goes, well, what happens if I give them a ten per cent SPIFF on this particular brand? That's the kind of business for it.  

 

Adam Thier 0:11:35.2: 

So at the bottom most layer we have Polaris, we have Forecaster, which is a complete ML engine. We have Optimizer, which is our Deep Algorithm Library. Yesterday we added a new one, which I'll show you in a minute, Syrup, which is a retail ML. Part of our very successful supply chain business. I mean hugely successful supply chain business. On top of our calculation layer, we now have the ADO which includes the common data model. So at the end of the year, when all 25 apps are out, they all share a common data model. So there are over a hundred common dimensions across all 25 applications, and when the partners come in to do it, they're going to be sharing that common data model, and the planning flows are pre-built. So as you see here, demand becomes sales, becomes revenue, is all pre-built flow across those modules. So all that stuff you figure out by yourself is now pre-built in there, and that common data model, that common data ontology, allows you to extend it across the business, so it's all metadata management. Above that, we have the apps layer. If you have apps, if you don't have apps, I'd love you to have apps if you don't have apps. I'm happy too, right, it's all good. I'm just the sales guy [?so I'm not really happy right now 0:12:54.4]. But you know, you do what's best for you, right, because you've got to. Then on top of it we have our agent layer. So analyst agents, and then the CoModeler agent, which I will show you in a minute.  

 

Adam Thier 0:13:09.5: 

So yesterday we did an acquisition. We did an acquisition last year on consolidations that's going very well, and now we have an ML specific to retail. So it's really good at replenishment. It's got some killer brand name customers using it, and that's what you're going to see with Anaplan. If we go back here, you're going to see a lot more of these ML engines there, but one thing I can assure you, one thing I can absolutely promise you, the original Anaplan that you all grew up with, where it had a modelling consul and a calc engine, that's never going away. Everything we're doing is wrapping and building around that, but if you want to run home to your calc engine and your model builder, and just build a new model from scratch, that's always going to be there. That's why we're successful, is you can always go, when the CFO comes in and goes, 'I need this by this afternoon', you can still do it by this afternoon. So we're wrapping all these layers of enterprise, stability, and relatability, but that core Classic engine, that core Polaris engine is always going to be there, and that I absolutely promise you. What you'll see is more innovation that pushes these things out, so you won't have to do a bunch of time spending it how to do retail ML, to do replenishment. That will be part of the platform. So allocation and replenishment, buy optimization, markdowns… It's a pretty comprehensive set of technology. We're very excited to have these people aboard.  

 

Adam Thier 0:14:49.6: 

So let's talk about our architecture. Our architecture is one hundred per cent open. So we talked about AI. AI agents on anybody's agenda? Yes. It's all about AI agents, and AI agents are those helpers. You know, you read all this stuff - the word agentic didn't even exist a year ago. I'm not even sure I like the word agentic. It's a made-up word. But at the end of the day, these are the things that are going to help you do your job, sometimes autonomously, sometimes supervised. It doesn't matter. The point is there are going to be a lot of them. So our entire agentic framework is built on what was the Google, but it's now the Linux Foundation. It's open. It's being supported by literally everyone, and what happens is, we consume the data, it goes up into our core, which wraps our calculation engines, forecast and the other things, and then the agents sit above it, and we are an open AI architecture.  

 

Adam Thier 0:15:47.0: 

I'm going to show you AI today using ChatGPT. I also use Gemini. I also use Claude a lot, and it's interesting to watch how, as the new ones come out, they leapfrog each other all the time, and that's going to be important for you, because the same thing's going to happen with you. You're constantly going to be leapfrogging. We have an obligation on our part to make sure they all work. But the point is, any agent you build an Anaplan studio, agent studio, or any agent we give you, it's automatically interoperable with any other agent built in the Linux Foundation framework. So that will be any agents you get from ServiceNow, any agents you get from Salesforce, any agents you get from Claude, any agents you get from OpenAI. Literally, they call them gardens. Google calls it their Agent Garden, where you go and find all the agents, and tell them to work together. So Agent to Agent Integrations, the open standard and protocol. It was introduced by Google. We're a hundred per cent onboard with that. Like I said, Amazon, Microsoft, Cisco. Everybody's using this. So we're right down the middle. 

 

Adam Thier 0:16:53.7: 

Model Context Protocol Support. We're going to have MCPs. That's kind of the secret sauce to make what happens happen - I'm going to show you here in a minute - to make all this work. It's all in the Model Context Protocol, and we're going to have an open support for Model Context Protocol. So again, has anybody here's company already done a deal with one of the big LLM providers, like Chat? Gemini? How many people are using Gemini? Claude? Right. What's happening is just like we went and bought capacity from Amazon, and we bought capacity from Google, these - our companies and Anaplan just did this - we bought a major chunk of Chat engine, and it's become our standard internally. Then Open and Extensible because there's - we are not even at the end of the beginning of AI, we're at the beginning of the beginning of AI, and it's all going to keep changing.  

 

Adam Thier 0:17:53.3: 

So here we go. Flying without a net. Can everybody see my slide? Look at that. Nothing up my sleeve. We have a basic Anaplan model here. Let me show you. Empty. Empty. All empty. So if - you'll see this thing call CoModeler. Let me click on CoModeler. So I've prepopulated a prompt here, because it's such a pain in the ass to type then in. What I'm going to do here is I'm going to - I asked it, 'Please build me a transportation planning module in Anaplan, using the Disco methodology. Please fully annotate the notes fields to provide adequate documentation. Basic functionalities should include Freight Management, Freight Cost Calculations' - and notice I'm spelling things wrong here - 'Freight Auditing and payments, Transportation Mode Selections and Shipment tracking and visibility. Why did I pick this? Anybody ever work in transportation? Anybody familiar? It's hard. It's a lot of complexity. So we're just going to click go. And it's analyzing my query. When I ran this yesterday it took 16 seconds. So I'll just sit up here, and tick tick. No. It's actually - right now, it's going out, understanding transportation planning, assembling all the things related to it, analyzing the Disco methodology, and look, oh my goodness, there are the requirements. Freight Management, Freight Cost Calculation, Build Progress. It builds it all out. I've outlined my key requirements. Should I begin executing the build? Sure. Please execute the build. So what it's doing now, it's thinking. At least it says it' thinking. It's opening the model. We'll give it a second. 

 

[Pause 0:20:07.2 - 0:20:17.9] 

 

Adam Thier 0:20:17.9: 

Look at that. So what we've built here is a world class, best in class model builder. Now, before you go, what am I going to do with the model builder? Right. What am I going to do? The answer is, it's not going to get it right. It's going to get it mostly right. It's going to do a good job, but it's not going to do as good a job as you. Right. It's not listening to the requirements, it's not understanding what's going on. But it is building a really good model, and I use this thing all the time. When I was prototyping what I was going to say here today, and I was like, what should I take to these people? Let's do something interesting, transportation, I built out a full transportation planning module. Not just this, but I went into driver scheduling and everything else, while I was sitting there watching the Jets lose on Sunday, and I went back and checked how much it cost me. It was $12 in Chat charges. That's what it cost me to do. So it adds up, you know, if you add in time, that took an hour. But the point being, it's building out, it's building the formulas, it's building everything you need. What we have seen with the customers, like yourselves, is - oh, there really is nothing up my sleeve. What we've seen with the customers like yourself is, now you can go and sit down with your business users, right? You can eliminate the backlog of planning. You can start doing way more interesting and effective things with this, but no unwashed person, no non-planner is going to really understand what's going on here. So this is going to have a huge impact, and in fact that impact is starting now.  

 

Adam Thier 0:22:02.3: 

So if you are interested in CoModeler, we are starting the betas next month. Now how the betas going to work is, we're going to come - [?Joe] and I and our teams are going to come and do a hackathon with you. So we're going to sit down with your team where you can come to our office, or we'll come to your office. We'll sit down and do the hackathon, so that you understand how to prompt, what's the - prompting was like - he was yelling at me because, Adam, it's not populating the notes fields, and I'm like, please populate the notes fields. It did nothing. I figured out the word was annotate. This is no joke. This is a true story. Until I picked the word annotate, it wouldn't populate the notes field. But as you can see here it's populating the notes field, so it's documenting the model. So we're going to do that, starting in November, and it's pretty bulletproof, otherwise I wouldn't be up here showing it live. So this is the latest and greatest.  

 

Adam Thier 0:23:02.9: 

Now, it's pretty cool because we've already taken your average Anaplan implementation. When I got here, going on almost three years ago, the average Anaplan implementation was 46 weeks. With applications it's about ten to 12. With this, we don't know what it is, but one of the things this will do, is you take an existing model and say I want to add this to an existing model, it can actually update, understand the model, and add the stuff without breaking it. We're working on it now for Polaris migrations. Can you migrate this Classic model of Polaris? All this, like I said, we're at the beginning of the beginning of the beginning with this. So I think we've seen enough. Let's see what needy questions CoModeler has for me. 

 

[Reads from website 0:23:51.9] Oh, pretty much done all the way. It even asked me to sign off. It didn't tell me where to sign off, but - and you can save these, obviously you can see these chats, so you know what to do next. But right now, it's a huge productivity boost to the model builders. Right. Can we go back to the slides? 

 

[Unclear interjection 0:24:14.2 - 0:24:22.4] 

 

Adam Thier 0:24:22.4: 

Yes, and if you're interested in being part of the beta, get your name in fast. So traditional implementation, six months, 46 weeks. That's what it was. With CoModeler it's going to bring down - we're expecting about a 50 per cent boost in productivity, is what we're seeing. You've still got to do the [?U axis 0:24:41.5]. We're working on it, automatically building [?U axis 0:24:44.3]. You've still got to go through reviews, but the best part about it, like I said, some of our European customers that were very early with this, the model builders go sit with the business users, and build the model side by side, and it has a huge, that feedback loop, and we didn't expect that, but these people - let's go sit with the business user and protype this stuff in real time, and that was the Polaris model by the way. They built that in Polaris. With applications, right, CoModeler is going to be the extending method for applications, because it understands the applications. So now you want to modify your application, it will extend the application we provided, and it will do it in a non-destructive way, so the system will still be upgradable. So when the next upgrade comes out it hasn't broken it. It doesn't make those mistakes that I make all the time.  

 

Adam Thier 0:25:37.9: 

Then finally we get to the bottom. We're talking about Anaplan implementations dropping dramatically, which allows you to have more Anaplan, to get more data from the agency organization, to make those better planning decisions. That's what we're trying to do here. It's not about, you know, making Anaplan cheaper, it's about making you better by having more Anaplan. What do you get in transportation? What does it cost, right? All of a sudden, your fuel prices go up, your transportation lane's changed. These have material impacts on your business. Now it's part of your Anaplan model. So you don't have to run off and go, well, what's our transportation cost's going to go up now the diesel's, you know, gone like this? Or they added an excise tax in California, again, for diesel, right, it can model it right away. It's all happening in real time. So ultimately, we're talking about very reduced kind of value with Anaplan.  

 

Adam Thier 0:26:39.2: 

Again, for the partners, now the partners can spend more time on helping you transform your business to what it would be or should be, had you known these things, right, reinventing, reinvigorating your processes. So we're making AI work for everybody. We've got the analysts for the planners and the analysts. We've got the anomaly detection, and we've got the analysts also for the executives, and then we've got Agent Studio and CoModeler for model builders. So we're touching everything here, and again when you look at Anaplan and go, well, how come Anaplan was able to build off this model builder compared to everyone else? The answer is, because deep down inside, we have linear algebra [?branch in 0:27:23.8]. It's the precursor to AI and that's why all this works. It's one stack, and did we get lucky? Were Michael and I really smart 26 years ago when we did this? No. We were guessing. We got lucky. At the end of the day, without having a linear algebra calculation engine, because guess what, my friend Chat really understands linear algebra, because he was built with it, it all works really well. 

 

[Video plays 0:27:55.9 - 0:29:16.4] 

 

Adam Thier 0:29:16.4: 

Again, so that's on the, how do I help the average employee across the company. It's right there, and it's fully auditable, and fully traceable. So if you ask it to show you that data, it will take you to the model. Make sense? It will show you access so the providence of the data is maintained, and the auditability of the data is maintained one hundred per cent. So you can go, okay, it told me this, let me check and make sure it's right. Okay, it's right. Move on.  

 

Adam Thier 0:29:52.4: 

Anomaly. This is our next big thing. It manages the data quality but also handles those anomalies. That whole thing about the port congestion. Well, what happens when the box doesn't come? What do you do today? You go and you look up the UPS number, you go to the UPS site, you put the UPS number in there, it tells you when it's going to come. How about an agent that automatically tells you this is going to be late, here's where it's going to be at, and here are your three options for dealing with this. That's what anomaly detection is. Or looking at trends for seeing pricing go down or seeing demand go up in this area. What do we do about it? So anomaly detection and running the scenarios automatically. Remember it said, go and update the plan. It could have just said, I've updated the plan, do you want me to commit it? That's where agents are going. This is a big deal, especially with the amount of data we're taking in. Like I said, every customer is bringing in hundreds of millions of rows. There are 2.14 million Anaplan models in production today this very instant in 22 data centers around the world. There are 3.3 million integrations running to Anaplan every single day. We have 2500 customers running 3.3 million integrations. Think about that. We are adding 70 gigabytes of new data to our estate for our customers every hour. We currently run 7.4 petabytes of data in memory at any point in time. The amount of data coming into Anaplan is crazy. The ability to detect anomalies in that data before it helps you make bad decisions is critical.  

 

Adam Thier 0:31:47.1: 

The next one's Workflow. As Anaplan stretches across the organization, get it in the right place, the right time to the right person is critical. So we'll have a Workflow Agent to build those workflows. You can build the workflow right from analyst, and they'll say, send this to these three people and get their approvals. So workflow is becoming entirely agentic. So remember, we didn't even have workflow two years ago. Now we've got AI-based workflow. So key takeaways. This is hugely transformational. Everything that you interact with Anaplan is the same but better; much more data, much more scalable, much more quickly, and then the AI is there to help you make those estates much bigger, to make more models, to manage the data, understand what's going on. You wouldn't be able to manage these estates without AI to help you. You just couldn't. It's beyond human comprehension.  

 

Adam Thier 0:32:55.1: 

AI is central to our product strategy. Again, all the way back to that linear algebra engine. It is central to what we do because the world is changing. It's almost - you know, we've been talking about this - I don't know, we were doing knowledge-based systems in the '80s and '90s. We've been talking about AI a long time, and finally it kind of is here. I was Bayesian. I was doing Bayesian inference in early AI systems 25 years ago, and it was hard, and it was clunky, and it really was not - you know, you could build like a cool thing that could do a thing, but it wasn't like it is today, and if you've seen there's a lot of pressure to do AI, and the beauty of it is, you can go home to your organization and go, there's a lot of really actual value AI in Anaplan. It' not AI washing, it's not just some - it literally is at the core of how I build and extend Anaplan though my organization. Like we said, AI requires an enterprise scale data platform. You look at the companies that are truly the most successful companies in AI, like Palantir. It's all about the data. Ninety per cent of what they do is the data management. You look at Databricks, you look at Snowflake. It's all about managing the data. You think about the amount of data your company is producing every day. Well, it's got to get fed back in and used. So that's kind of the end of my story. Do I have something else? Oh yes. Visit the Anaplan product experts in the customer lounge. Come and see the products. I'm going to take questions.