In this interview from Appian World 2026, Jake Rank, vice president of product management at Appian, joins theCUBE's Dave Vellante to discuss how organizations are embedding AI reliably into operational workflows through purpose-built process platforms. Rank walks through a series of product announcements, beginning with significant enhancements to Appian's document processing capabilities — an area that surged in customer adoption in 2025. He explains how AI-generated configuration recommendations can push document extraction accuracy from 80% to 99%, giving enterprises a structured path to value rather than a frustrating cycle of ad hoc AI experimentation. Rank also details how Appian's expanded agent offering now supports continuous human feedback, with one internal test showing an agent jump from 75% to 95% accuracy in just 30 minutes.
The conversation also explores Appian's data fabric — a read/write virtualization layer that unifies enterprise data across sources without requiring migration or re-platforming. Rank contrasts it with data warehouses and materialized views, underscoring that data fabric is an operational system capable of powering real processes, not just reports. He details a context layer that encodes semantic meaning so agents know which records to query, which relationships to follow and which values to filter — eliminating guesswork and enabling reliable agentic decision-making. Rank also addresses Appian's adoption of MCP, which opens its data and process assets to external agents and AI systems across the enterprise ecosystem. From managing model drift through continuous feedback loops to mixing deterministic and agentic workflows for regulatory and cost efficiency, Rank provides a practical roadmap for how organizations can embed AI into core operations without sacrificing control.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Appian World 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Appian World 2026
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Appian World 2026.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Appian World 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Appian World 2026
Please sign in with LinkedIn to continue to Appian World 2026. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Jake Rank, Appian
Dave Vellante hosts a conversation with Jake Rank, VP, Product Management, Appian, at Appian
World 2026 from the JW Marriott Orlando, Grande Lakes in Orlando, FL.
In this interview from Appian World 2026, Jake Rank, vice president of product management at Appian, joins theCUBE's Dave Vellante to discuss how organizations are embedding AI reliably into operational workflows through purpose-built process platforms. Rank walks through a series of product announcements, beginning with significant enhancements to Appian's document processing capabilities — an area that surged in customer adoption in 2025. He explains how AI-generated configuration recommendations can push document extraction accuracy from 80% to 99%, giving...Read more
>> Hi everybody. Welcome back to Orlando, Florida. We're here at the JW Marriott, and this is theCUBE's live coverage of Appian World 2026. We've been talking a lot about process, bringing probabilistic and deterministic AI together and really solving hardcore business problems, tough business problems. We've been calling it East Coast AI. This is coined by the CEO of Appian, Matt. Jake Rank is here, is the Vice President of Product Management at Appian. And we're going to get into where we're at in 2026, some of the announcements that were made at this event. We've talked a little bit about data fabric. We can double click on that and agentic. Jake, thanks for coming on. Good to see you.
Jake Rank
>> Yeah, no, it's great to be here.
Dave Vellante
>> All right. So let's do a rundown. You guys made a bunch of news this week. We heard some of it in the keynote. Lay it out. What's the big news? What should we know about?
Jake Rank
>> Yeah. Well, I mean, we've put out a lot of new features, a lot of improvements to existing features. We're really doubling down on our overall goal to make applications reliable and get results from AI. So one of the big innovations was we've doubled down on everything that we did around document processing, which is an area that really exploded in 2025. We saw huge amounts of results from customers there last year. And so we've dramatically improved, added more capabilities to that offering, and we're expecting to see big results from that in 2026 as well. We also, of course, enhanced our agent offering, expanding what we can do in terms of reliability, expanding the tool sets that it can take advantage of, and opening up new agentic collaboration with humans and agents working together. And then as you said before, data fabric. Data fabric has always been a core foundation of what Appian's product has to offer, and it's a foundation upon which we build a tremendous amount of value with AI. So we've enhanced that, expanded the scale, expanded the amount of data types that you can pull in, including documents. And we've started to open it up to the ecosystem more so that people can use Appian in new ways with agents in Appian, agents outside of Appian, just opening us up and giving us more capability. So many of those different things that accelerate how people can use Appian to automate their business, but also making it faster to build, right? Modernize applications, use AI to build Appian applications and get to the outcome faster.
Dave Vellante
>> Let's go through some of those and try to connect the dots for folks. So start with document processing. You said it really started to take off in 2025. I presume you did some document processing prior to 2025, intelligent document processing, document processing. Now, it's like super intelligent document processing. So presumably, generative AI had an impact on that whole space. Take us through sort of the changes in the market and how you guys responded to it.
Jake Rank
>> Yeah. I mean, it's been very interesting because we have been working on it for a number of years. I think maybe six years. It was 2020, I think, when we released the first version of Document Center. And the technology has obviously changed a lot since then. So I think one of the things that we've seen that's opened it up, GenAI has made more powerful with easier configuration. But honestly, under the hood, all the technologies are still available. And there are places that you want to use computer vision or machine learning for certain use cases. And in Appian, you can easily select all those different technologies. We're making it really easy to get the right tool for the right job when you're processing different types of documents. Now, one of the things we just added now was more support for Excel. So large Excel files, which every company has and needs to really process in an automated way, being able to apply AI to that is now going to help us automate even more tasks.
Dave Vellante
>> Like what? What can I do as like a power? You're talking power Excel users, right? With big files.
Jake Rank
>> Sure.
Dave Vellante
>> What kinds of things can I do today after these announcements that I couldn't do before?
Jake Rank
>> Right. So you might be downloading Excel from a legacy site, somewhere where you don't have an API to integrate with those systems, but they just post an Excel and it's got 50,000, 100,000 rows in it. Maybe it's got some macros, maybe it's got split cells, it's got multiple tabs, right? Just the kind of Excel that really gets business done, but is also kind of complicated to look at.
Dave Vellante
>> You're going, "Ugh."
Jake Rank
>> Yeah. And I mean, honestly, yeah, do you want to sit there and copy paste a bunch of stuff to find the exact row that matters? No, you want to digitize that in a way that's useful and that you can bring into and associate with data fabric. And now, you've got it, you've unlocked it. You've associated it with all of the other enterprise information in your ecosystem and agents, humans, everybody can use that as part of the processing.
Dave Vellante
>> Interesting. I mean, my personal experience with such tools, not your tool, I haven't used it, but with sort of Excel, Copilot assistance, I've found them to be clunky, not that accurate. To quote Marc Benioff, Clippy-like. And so I presume you guys were well aware of some of those challenges.
Jake Rank
>> No.
Dave Vellante
>> So how did you address that? What's the experience like for customers?
Jake Rank
>> Yeah. Well, I mean, to think that something's going to come off like Clippy or something like that, it's very naïve, right? It's not going to get you the real results that you need. So even with Excel, with other document formats, what we've done is we've built in the checks so that you can build test sets and you can evaluate those test sets. When it doesn't meet your needs, when you're maybe getting 50%, 70%, 80%, not getting to be reliable, we look at all the failures and actually use AI to automatically generate recommendations for how you should tweak the configuration so that you, even being a non-expert, can sit down and very rapidly go through that feedback loop to go from 80% to 90% to 99% accurate. That's actually a number that we see pretty often with our customers is 99% accuracy and also the confidence to go forward with accuracy and actually go straight through processing.
Dave Vellante
>> This is super important because, again, my personal experiences is when you get frustrated with the AI, you don't want to go in sell by sell and figure out exactly what's going on. So you try to get the AI to do it. It's not working. So finally say, "All right." It's like, "I'll just do it my ... If you want it done right, you got to do it yourself." And you go in and it's painful and you just wasted the time with the AI. It sounds like you guys were thoughtful about that and have had great success, so that's-
Jake Rank
>> Yeah. I 100% agree. I ask you, my own personal experience mirrors yours. You sometimes invest time and you don't get the payback. We are not asking people to use AI on its own. We are asking people to use features that are built with AI as part of the backbone. But the fact that we've built all the structure around, built the feature around it, you're getting the value, you're getting an assured path to value because it's not just ad hoc AI, it's actually a feature. It's an intentional way of solving a business problem that happens to use AI.
Dave Vellante
>> So it might be do this at some point, but it might not start with some superficial do this. It might start with, "Where should I be focused? How does this work? What does this mean?"
Jake Rank
>> Sure.
Dave Vellante
>> And then maybe I'll tell it to act on my behalf at some point when I trust it.
Jake Rank
>> In this case, we're not even asking you to make that decision, right? You're uploading a file. That's your intent is to extract from that file. You don't have to prompt it for anything. It already knows the intent. So you're taking a whole category of possible mistakes off the table and that's what leads to the fast results and the reliability.
Dave Vellante
>> How does it know the intent? Because you've got application logic and process knowledge that's being injected into the system.
Jake Rank
>> Right. We've looked across all the different customers and the kinds of challenges that they're facing in their workflows when they're trying to process documents. And we've built specific features that address those needs. So now, that you just use a feature that's designed to extract from documents, other features like agents, other features can do other things, right? Generate text, summarize an email, those are all different business tasks. And we've built features that are fit for each of those tasks. So you're not just using AI, you're using a feature that's designed to solve your problem.
Dave Vellante
>> That's pretty dope as the young people would say. All right, let's talk about agentic collaboration. What is agentic collaboration and how is it different than human collaboration?
Jake Rank
>> Yeah, sure. So we released our first version of agents last year, and now we're continuing to invest. We've released new capabilities for agents, just like with DocCenter to add reliability and make it easier to get to results. So similar ways that humans can very easily add feedback to their agent runs just as easy a thumbs up or thumbs down. Now, you can make your agents better. You can get recommendations and it'll automatically tell you how you want to modify those agents. Now, we often see agents be put into a task flow. Maybe a human is doing a task today and you can replace that human or augment them with an agent. But of course, there's also going to be circumstances where you want to have a human still be in control, right? Agents can be in control of some tasks, humans can be in control of other tasks. In those cases, we've now also allowed you to inject an agent right into the screen so that you can have your task and your agent that helps you complete that task. And that's really important in today's business because you're seeing the loss of institutional knowledge as there's been workforce turnover. And so you have new people coming in, an agent can actually bring all that context of years' worth of institutional knowledge to bear even if that person has only been on the job for a week.
Dave Vellante
>> I want to come back and ask you about the thumbs up and the thumbs down, because as a user, I'm often skeptical of it like, "Okay, should I spend the time?" If I do spend the time, I'm just helping Google get richer, open AI scale, maybe I'll let somebody else ... But they have the scale. My point is they have the scale and even superficial feedback, they can extract some nuggets and at scale they can come up with the answer. Do you guys use a similar technique? You don't have obviously the scale of Google. Or is it more you've got some secret sauce to interpret based on whatever actions were taken, or is there an enticement because it's enterprise, people care? How do you take the thumbs up and thumbs down and make it non-superficial?
Jake Rank
>> Sure, sure. I mean, I think that large vendors do have scale, but they don't have scale on individual customer processes. Our customers are actually the experts on their processes and their tasks. So they're the best ones that are qualified to actually give feedback. It's a very focused approach on that specific task. Again, going back to the overall strategy of using AI in a constrained way so that you're actually solving a specific problem. And by giving feedback on that specific problem, you're actually just doing what people in the business are already doing. They're already getting the results of these actions and they're already saying, "This is good, this is bad." Now, we're able to leverage that feedback. So we're actually improving the way that they can increase that feedback loop. Instead of sending an email, they can just put it right into the system and the system automatically processes it to create recommendations for improvement.
Dave Vellante
>> So you need depth. Obviously, you've got depth, process depth. You do need sufficient scale, but it doesn't have to be Google scale, nowhere close to that. And you probably need a little bit of time and enough users interacting with the system, and then you can kind of get to that very high level of accuracy. What's been your experience? I mean, it's still early, but what kind of curve are folks on? I can see with our own AI, I'll come in, maybe for some reason I haven't used some tool for several weeks and I come back and it's really, "Wow, this is really good now. It used to not be so good. Something must have changed." What are you seeing in terms of the time horizon of that improvement or the shape of the curve of that improvement?
Jake Rank
>> I mean, it honestly can be quite fast. So when we were developing some of our agentic feedback mechanisms, we actually saw an initial agent that was 75% accurate, go up to 95% accuracy with just 30 minutes of feedback. So it's quite rapid actually, and you don't need a ton of instances. Now, of course, as you deploy that into production and business changes, new types of partners are onboarded, new customers, trends in the world change. So you've got to keep that accuracy high and the consistent feedback also helps refine and keep feedback, keep the accuracy high over time.
Dave Vellante
>> Yeah, because there is this trend in LLMs they will degrade over time if you don't ... It's like all of us, you got to keep in shape.
Jake Rank
>> Yeah. We call it drift, right?
Dave Vellante
>> .
Jake Rank
>> If the business actually changes, and if you leave your model behind, you're not changing with the business. So keeping fresh feedback helps you stay modern.
Dave Vellante
>> So agentic collaboration, just to put a finer point on it, is the collaboration between agents and humans, so that's like-
Jake Rank
>> Right. We've made that new. You now have it directly in your interface. So when you're doing a task in Appian, you can have an assistant right there to help you do it.
Dave Vellante
>> Okay. Data fabric, that term, I think Gartner coined the term data fabric. There was a little battle between Zhamak Dehghani's data mesh and Gartner's data fabric caused a little confusion in the market, but I think data fabric is more considered a technology. Data mesh is more of sort of an organizational construct, and so that's sort of been cleared up. But data fabric is this way to essentially virtualize your data without necessarily having to move the system. Tell us more about data fabric. What's the history of data fabric? Why is it so cool?
Jake Rank
>> Yeah. I mean, Appian's been working on data fabric four years, even back before some of the official terms we're calling it-
Dave Vellante
>> So maybe you coined it, maybe not. Maybe it was called something else. Anytime we can give credit to Gartner, we love it because they need credit. Anyway.
Jake Rank
>> So data fabric has become one of the essential foundations of Appian. And it's really powerful, because what it does is, like you said, it brings together information from across your enterprise without the drama of data warehouses and actually having to re-platform your information. Data fabric allows data to live in the sources of record, but be virtualized in a layer that allows you to connect and relate information across different sources to secure it at a row or column level, and then to turn around and expose it to all the other capabilities in Appian as well as outside of Appian. So every screen can now be a 360 degree view no matter where the data lives in your enterprise. So as a foundation, it's been incredibly powerful. And now with agents, it's even more so.
Dave Vellante
>> What is the kind of secret sauce of data fabric? Is it some kind of data virtualization technology? A second question, is there a harmonization, data harmonization component? I'd love to drill into that a little bit.
Jake Rank
>> Yeah. Well, I think we saw a problem and then we went and found a solution and it was hard work. I mean, it was a lot of engineering, a lot of understanding what our customer's problems are, a lot of working with customers over time, but as we've evolved it, we've managed to create something that I think has been truly successful and that is very widely adopted amongst our customers. And again, now that we have that foundation, we're able to build on top of it because everything else that we do benefits from that foundation.
Dave Vellante
>> Kind of a Colombo question for those of you who are old enough to know who Colombo is, but can I achieve the same thing? I know I can't, but explain why. We'd like materialized views. So explain what's different, because data fabric is a fundamental foundation and I'm presuming there's read, write capabilities, but maybe not, but materialized views for years were kind of the way in which we sort of virtualized our data. Explain the difference.
Jake Rank
>> Yeah. Well, there's a number of ones. You actually hit on one, right? I mean, you're not having to move the data, which materialized views is great if you have access to all that information in that particular database engine, right? But now, you can relate information from all the way across the enterprise, even from web services, from public data sources, you can bring it all together in data fabric. And then like you said, it is read, write. So unlike a data warehouse, unlike maybe a data lake, actually, it's an operational system. So you're actually able to build not just reports, but also processes that do real work using data fabric. And I think that that really sets data fabric and specifically Appian's data fabric apart from what a lot of other vendors have.
Dave Vellante
>> So irrespective of the data store, the storage format, the physical location, cloud, on-prem, edge, doesn't matter?
Jake Rank
>> Mm-hmm. We have many customers who are using VPN to on-premise data systems, even if they're in Appian cloud. So there's many ways to bring the information in. And once it's there, you have a common layer, a common foundation to build agents and UIs and other services for your enterprise.
Dave Vellante
>> Okay. But for me, to go, Jake, from that data ingestion to being able to have an agent trust it, there has to be some kind of, call it semantics context. I call it harmonization. Sometimes people confuse it with ... Semantics can be at a BI platform and it doesn't capture the entire where I'm going at. But so is there a harmonization capability that's inherent there? Is that evolving? I mean, I think about Palantir Ontology as sort of the thing everybody talks about, but then of course, you need forward deployed engineers to make all that happen, but still very powerful. I'm learning here that you guys have a very powerful technology. Is there a harmonization component to it?
Jake Rank
>> Yeah, and there is. I mean, of course, you don't need forward deployed engineers all the time and it's also much more operational than maybe some of these services that only do a lot of reporting. And we've call it the context layer and actually it applies to data fabric. It applies to all of the things that you build in Appian, but the idea that you have metadata, you have a description, you have a title, you understand the meaning of the data. So semantics, right? You do understand what the data represents, which means when an agent looks at that information, it understands how to use it effectively. It knows which records to query, what relationships to follow, what values to filter by. All of that is encoded in our context layer. And again, data is really powerful, but it does the same for processes, business rules, documents. All of what you build in Appian has the context layer that allows us to use agents effectively on top of it.
Dave Vellante
>> So this is what we, in our parlance and theCUBE research, we call this the system of intelligence. Some people maybe think of that as legacy. We don't. We think there's got to be this intelligence layer. Some guys on our team call it a cognitive surface, but it's intelligence that interacts ... Well, let me ask you, what role does the LLM play in all this?
Jake Rank
>> Well, I mean, certainly LLMs are what I've unlocked so much, but if you don't put an LLM or a generative AI into a system that does real work, it's not really doing much. So the LLM is responsible for understanding the natural language, which is what's the context layer is providing and then reasoning about the decisions that it wants to take. So it's an important part of the system, but it's only one part of the system.
Dave Vellante
>> But people realize that taking in an LLM and sticking it in front of a database, even if you vectorize it, it's like, "Okay, now what? What can I really do with that?" And so you've got to have this sort of what we're calling a system of intelligence to be able to harmonize that data and then serve it up to agents so they can be trusted. Okay, cool. That's very helpful. Okay. Now, the ecosystem piece of that, so that's opening up data fabric to the ecosystem. Explain that, how you do that. MCP is involved?
Jake Rank
>> Absolutely. Yeah. I mean, we've always been a good player in the ecosystem and Appian goes into complex enterprise environments that have lots of systems and it's important for us to be able to work with those systems. So we've always had strong integration story. Now, we have MCP so that we can take all that data and the context layer that's providing it, make it available. If you want to use it from your other systems using AI or other agents, you can. And at the same time, we've enabled our own agents to use MCP tools. So as MCP or its subsequent technologies that may come after MCP come, we're going to be able to keep using all of those same tools and all that value that you've built with the tools for data, all the value that you've built in Appian with the tools for processes are going to be available for you in your enterprise to take advantage of whether you're using Appian's AI or other AI.
Dave Vellante
>> So that means the entries and the exits, entry into and the exits from your system are open. You obviously have your own IP, you're not giving that away, but you're letting people in and you're letting people get out.
Jake Rank
>> Well, and we're still a platform where you can come in and build a process. I mean, even in this world of agentic processing, there's still certain things that you want to have happen in a very precise way, whether for regulatory reasons or speed or cost, and we provide those tools as well as the agentic tools. So mixing deterministic processes and agentic processes together lets you pick the most appropriate way to solve any problem.
Dave Vellante
>> Yeah, you don't always need generative. In fact, sometimes you don't want it in there. I know I got to get from point A to point B. This is the most efficient path. I know it, it works. Boom, make it happen. All right, Jake, hey, thanks so much for coming on theCUBE and going deeper-
Jake Rank
>> Yeah. No, thank you very much....
Dave Vellante
>> on announcements. It was very useful. Appreciate it.
Jake Rank
>> Yeah. Appreciate it. Thank you.
Dave Vellante
>> All right. Keep it right there. We're winding up day two. This is Dave Vellante for Alison Kosik and the entire CUBE team, Appian World 2026 from Florida. We're gonna be right back, right after this short break.
Dave Vellante