Hosted By
This episode of the Coffee With Digital Trailblazers was sponsored by Appian.
The event is hosted by Isaac Sacolick on LinkedIn Fridays at 11 am ET. The event attacts digitial transformation leaders, from CXOs to team leaders, who learn from experts on driving change in their organizations. Every week we explore a topic and share lessons learned, and all are welcom to attend. Visit https://starcio.com/coffee/next-event which will redirect to the next event.
Recordings of the events are available to StarCIO Digital Trailblazer Community members .Review the community’s advisory, leadernng, and leadership programs.
Isaac Sacolick:
Mike, I’m glad you’re here. We can get started now. Welcome everyone to this week’s Coffee with Digital Trailblazers. We are excited to have this week a session on why AI is boring in 2025 and how all of you can learn a little bit about how to accelerate your transformation programs. Today’s session is brought to you by Appian, a software company that orchestrates business processes, and I want to thank Mike Beckley for joining us today and being our guest speaker. Welcome John, Joe and Joanne to the floor. And Mike, we can give it a few more seconds just to get some more people to join. It’s usually like 1104 or five where I really get started with everybody, so everybody should be getting ready for in the US getting ready for your Thanksgiving weekends. I’m really excited. I’ll be doing some travel this week to go visit my son in Arizona, so if any of you’re in Arizona and want to meet up with me, I’ll be in Phoenix for a day and I will be in Tucson for a few days and happy to say hello to anybody who is around again. Today we’re talking about why AI is boring in 2025 and how to accelerate transformation. Again, today’s session is brought to you by Appian, a software company that orchestrates business processes. Mike, it’s so good to have you on stage despite our few hiccups. I want to give you a form first, tell us why you think AI is going to be boring in 2025.
Mike Beckley:
Yeah, thank you Isaac. So glad to be here and to get past our firewalls. Finally. So what’s boring about AI is that it’s going mainstream, but it’s how it’s going mainstream that’s uniquely boring. The most powerful and effective use cases for generative AI are simply not flashy and exciting the way they have been in the past year. Regenerative ai, we’ve all gotten to see how amazing it is at drawing pictures, at painting, at generating photos and images, and it’s incredible at that. It’s also incredible at writing stories for us, and anytime your nephew wants to hear a scary story, you can just go to generative AI and create one. And it’s amazing at those types of tasks. But where it has fallen down completely has been in real operational workflows. And now what’s coming in 2025 is the where generativeAI is going to be delivering the most value. Where we’re seeing already where it’s been making incredible breakthroughs in business systems is in the most boring part of them. And that’s OCR, that is scanning documents, that’s extracting data from PDFs that has completely revolutionized the economics of scanning and digitizing paper, and it’s the most incredible thing to see dramatic multimillion dollar savings, but oh my God, is it boring?
Isaac Sacolick:
Mike, maybe go a little bit deeper than that. We’ve seen document processing before, it’s been around for a long time. OCR and then can scanning in my invoices and pick out dates and numbers out of it. They can scan my contracts and again, pick up dates and numbers around it. What is generative AI doing above and beyond that?
Mike Beckley:
Yeah, that’s really the whole point is that OCR, we’ve taken it for granted because well, it kind of works, but also that it kind of doesn’t. And so to make traditional AI using deep learning techniques made it a little better. But it still was a consulting engagement where humans had to train the AI on every document and had to give enough samples. And if there was variability even within that, the AI was a little better than the old template during models. But the fact is it was still cost prohibitive to actually automate most of your documents because you can only focus that kind of attention and AI engineering on the most high volume, most repetitive documents. And so that left hundreds of document types out. So if you’re an insurance company and you want to be able to quote fast because the faster quote gets the business, most of the time you’re dealing with hundreds of possible agents out there all submitting their own unique formats.
And with traditional AI models, you might get maybe 60% straight through processing. And now with generative ai, it’s solving all of the edge cases. And I don’t mean that without being specific. I mean all is all. So all the things that made AI fail before, tables that span multiple pages, tables without lines between the words and numbers, the documents got too large. Whatever it is. Handwriting generative AI brings the context to the document that makes this really error prone issue. No more error prone at all. Your strength through processing rates go up to like 97, 98, 99, 99 0.9%. It it’s insane. And syns insanely great and how different it is when you apply general, it just obsoleted the entire OCR industry virtually overnight.
Isaac Sacolick:
And Mike, you mentioned insurance. What other industries and types of companies should really be taking advantage of this?
Mike Beckley:
Well, it’s the most regulated, most paper intensive industries because the regulations require you to submit so many forms. So banking, financial services of all kinds, investment banking, asset management, private equity, retail banking. These are massively paper intensive transactions. Insurance of course, property and casualty life, all of it. And then the life sciences world, if you are trying to conduct clinical trials and working with many different outsource labs and facilities that the paperwork requirements are massive and putting drugs through trials, it is just a very document heavy compliance heavy sort of process because human lives are at stake and the documentation is never going away. But now with generative ai, we can actually capture that whole long tail of unstructured data and tructure it, then the government. This is really going to be hugely transformative in government. You want to make a dent in how much manual labor still is required for the government bureaucracy to operate generative ai.
Attacking the paper problem in a new way is going to be really, really impactful. But also, I can’t emphasize enough, it’s the paperwork that matters. It’s not talking about generative AI replacing the human agent. It’s not generative AI that’s going to somehow make the decisions. It’s generative AI that in 2025, it’s going to actually be able to finally fulfill the dream of the paperless office, and that’s going to transform government. It’s going to transform banking, it’s going to transform insurance and life sciences to start, but wherever the regulations are heaviest, therefore the paper’s the heaviest, that’s where the impact is exponentially greater.
Isaac Sacolick:
I’m going to go around the horn. Joe, you’re giving us a thumbs up to the paperless office. I’m sure you have a lot of examples of that.
Joe Puglisi:
I saw this week, Isaac, an example of where exactly what Michael is talking about has come to fruition. It was a startup company showing an RPA solution, and we know RPA is good as long as nothing changes, right? So version 1.0 of RPA was very, very rigid. Well, when you bake in an AI underpinning and the RPA instructions can be delivered in context, suddenly you’ve supercharged your RPA agent. And even when the forms change a little bit, whether it’s a new column or something else that would ordinarily derail the RPA script, it can be more adaptive and maybe get 98 or 99% accuracy as opposed to 60 to 70%.
Isaac Sacolick:
Yeah, I like the term Mike uses around exceptions because, and that’s why I throw out the invoice processing example. RPA has the same issue, which is you sort of design the flow to the ideal path and cheer success, and then all of a sudden you start putting in real world examples. It just doesn’t work. And I agree with Mike, the ability for it to not only just handle the parsing issues but handle a lot of the interpretation issues that come up when you start using long form content. It is pretty amazing. Joanne, what are you seeing that’s boring that we can really deliver value out in 2025 around ai?
Joanne Friedman:
Well, to the point about RPA Gen AI is not the only way to do this, but agentic AI will blow the RPA out of the water because it does both. We designed the flow to take the standard, and to Michael’s point, all of the exceptions, and you can actually, because agents are multi-layers and they sense and detect their environment before they do anything else, they can immediately discern whether it’s the exception or the norm to speed up the process. And you can actually let them go off and do their thing in an even faster way than Jenny and I can do it because it’s pre-programmed to not only sense and detect the anomalies and the environment, but just the way the frameworks for Gen KI are being created. That is, I wouldn’t say RPA on steroids. I would say more like this is a new class of technology that has the capability, particularly to the point about paper intensive industries and sectors.
Think about a hospital and the amount of paperwork that you have to sign, even if it’s on a tablet, now you can contextualize the same piece of data in different flavors. So your chocolate, vanilla, strawberry around the same is the different contexts that can be applied in agent ai. This becomes that much faster as quickly as something like Claude or Hitachi, BT can do the tasks in the RPA, the agents can do them even 30 times faster with less power, less compute, et cetera, et cetera, because it’s a framework that tells ’em what to do and go off and execute without human necessarily having to be involved. Do you trust it? You will right off the bat, but after 10,000 mes, absolutely you will.
Isaac Sacolick:
Thank you, Joanne. So Joe brought up RPA, you bring up a agentic ai, let’s let John, how do we take boring use cases and make them value in 2025? And then Mike, I’ll come right back to you.
John Patrick Luethe:
Yeah, and thanks for having me up here. I completely agree with Mike that generative AI has been so powerful for translation for information retrieval. I think the things that are holding it back right now and make it boring, two things are, one is that we’re using general purpose AI often for very specific use cases and well, that can be helped by augmenting it with information. I think as we get towards more custom built language models and gendered AI that are built just specifically for the use cases we’re in, I think it’s going to get a lot more interesting. The other one is that I think a lot of the people using gendered AI right now are interfacing it through a website. They’re pacing stuff in and they’re getting things back or they’re using Microsoft Copilot or one of the other ones where it’s like the humans are the interface to it. And I think once you start stitching generative AI into land of business applications, that’s when management happens. And so I know Mike here at Appian, just the ability to have users build stuff with low code or build stuff with you have the development team built up with regular development tools and start stitching generative AI into really into the workflows. That’s when it gets really, really neat. You start saving time,
Isaac Sacolick:
Mike, they’re throwing you the softballs because I was at your conference earlier this year and I saw some great examples of that convergence between what you used to be able to do with RPA and maybe it fell off and you get into intelligent automation, you get into low-code capabilities, plugging it into machine learning capabilities. Now Joanne is introducing agents which you introduced back in June. Tell me about this convergence, right, this convergence of all these capabilities. What are you seeing businesses being able to benefit from all these capabilities when they’re brought under together in a platform like Appian?
Mike Beckley:
Well, the important thing is with the new technology like generative ai, can you actually use it in these highly regulated industries? And can you answer that question for a regulator? Can you answer it for your customer that you’re keeping their data secret and private and you’re not training an LLM on their private information and therefore risking leaking it and sharing it in some unanticipated way because the LM is out of your control, you can only predict what it will do. You can’t necessarily stop it from doing something unexpected. And so what Appian has done for the last year, and I think some of our competitors as well is really focused on this concept of private AI and making sure that generative AI can be installed directly into a process or a workflow and kept walled off from, you don’t have to train it on your data.
You are keeping everything within your security perimeter, the data space of your control, and you’re getting the power and benefits of generative AI and gen AI to make your RPA to make your processes, to make your low-code applications to make them actually intelligent. And so that’s again, the most boring part of this is now we have the security guarantees in government we call the FedRAMP guidelines. We got to comply with, we have FedRAMP approved generative AI in banking, PCI compliance and healthcare HIPAA compliance. We have these compliance regimes that we’ve been able to get approval for generative AI in and a deployment mechanism through our partnership with AWS bedrock to actually deliver generative AI and agent AI directly into all these workflows that have previously thousands and thousands of people inside of banks are still doing manual work just because there wasn’t legal compliance approval to use generative ai. And that’s all those barriers are falling down in 2025.
Isaac Sacolick:
So I’ve got private LLMs, I’ve got compliance for the major industries. Can you comment on what agentic AI is going to do for these industries?
Mike Beckley:
Yeah, so agen AI is the new trendy term and it’s being used in wildly different ways. What Joanne was talking about is interesting in some ways different from what I see in that’s that most of what’s going on with Egen is the AI companies are trying to chain together multiple actions within the ai. So having philanthropic demonstrated plug itself as an RPA bot, if you will, going and driving a web browser. So when people say eTech ai, they usually mean generative AI driving a web browser as opposed to a script in a bot driving a web browser. And that is very primitive and slow right now, and it’ll get a lot better. I don’t know how fast it will evolve in 2025, but instead the effective agenda AI right now is using a workflow engine like an Appian to provide a tool chain to the generative AI and therefore having the chain of reasoning governed by the process that the humans have built or designed. And then AI can be so much more powerful because it can perform whole activities like an underwriting action within the constraints and the common sense guardrails and the goals that have been set by humans and within the compliance boundaries of the process. But it is not just a chat and discussion where I ask a question, I get an answer, it’s do something for me, and it goes through multistep, it invokes many different data systems and comes back with an answer. For me,
Isaac Sacolick:
Mike, you’re being consistent with your boring use case because basically my version of an agent AI is bring the generative AI capability to the workflow people are doing in the platforms that people are doing them with the data in that platform, but also the data outside of that platform that we make accessible and bring that human in the loop to make them more productive, more smarter, bring capabilities to them that they didn’t see before because it’s buried behind a lot of data ahead. Joanna, I know you have some thoughts around this as well
Joanne Friedman:
To both of the points that were made. The beauty of agen AI is that it actually can do not only the process calls, but the actual programmatic functions features. You can build it. There’s a variety of different kinds of frameworks. So think about taking something like a microservice or a container and saying to using an agenda framework. And as I said, there are plenty of them where you can actually tell the agent, go look at all my back office systems that would normally be used in this workflow, an ERP or PLM or a CRM package, whatever it is, and now also fetch data or go and query against a different kind of database, call it a time series and put all of these pieces of data together in a context that makes it very usable for the user. So you’re getting the right answer in the right context on demand and you’re getting the opportunity to either add to it, which means it could be used for a training purpose or it could be used for a knowledge capture.
And over time as it learns all of the pieces of information, it then becomes sophisticated enough that you can say, okay, now do this for me. Do this for me. Go and execute and do it with the assurity that all of the data from all the different systems when they’re choreographed together in the right way are going to give you the right results. And it’s not about using things like mag around the gen AI to make it more less hallucinatory. It’s about actually go do this for me. So all of the mundane tasks that don’t require a tremendous amount of strategic or real thinking around problem solving can be done by agents and you can actually have one agent collaborate with another. So if you do have complex problems that we do in manufacturing, I can put my agents together to get an answer like here’s the root cause and do it in an incredibly fast way.
So depending on the role I have in my workforce, if it’s in a financial services business, if I’m on mortgage approver for example, as opposed to someone doing another part of the process like validating and verify, I can have agents do the validation and verification. I gave another piece of paper called my T four, my income tax return to make sure that I’m qualified. I can do the same thing in manufacturing with a bill of materials. Both those workflows and both those processes could be done automagically or autonomously by the agents. So what it’s doing is it’s freeing up workforce time to work on the more complex issues that really drive business value. And that’s what fascinates me about it. And to see these kind of capabilities embedded in pieces of software that will then be sort of smart out of the box, that’s going to change the face of business, not just the face of technology,
Isaac Sacolick:
This discussion around how boring AI is going to change the face of business. And Joan, you’re bringing up another key point around where agents are going to be very interesting, but what you’re describing is role-based agents and looking at workflow and saying, I’m a security expert, I’m a privacy expert, or I’m not, and how can I get some help around what to do here given the circumstance or where we are in our flow? Mike, I’m wondering if you can comment on that, on that role-based agent ai and then also, again, coming back from being at the Appian conference back in June, I think it was, I lose track of time. What intrigues me about your offering is the intersection of these capabilities with your data fabric. So maybe talk a little bit about that as well.
Mike Beckley:
Yeah, well, lemme just say this. What I love about Joan’s talking about is how important the workflow is, and it’s becoming more important than the generative AI models and the lms, the underlying large language models themselves. Because what’s happening is that the models are becoming commoditized. They’ve all become so good that even the open source alternatives are catching up. And we saw China release a model, an LM model, which may or may not be nearly as good as the American versions in this past week. This is a kind of rapid convergence on LLMs that is devaluing the innovation of the LLM and instead shifting where the value will lie to the application layer, to the workflows themselves and therefore to get value to create these different role-based agents you’re talking about, it’s all about do you have the right workflow, do you have the right tools that the LM can act on?
And then what you just asked about data fabric, this is our approach. This is, I mean, app is not the data fabric now they’re becoming data fabrics are becoming quite trend and popular. Most of our low-code process automation competitors have recently announced new data fabric technology. Just because we were one of beginning early adopters, it’s really become common to see because what’s a data fabric? A data fabric is a way to work with all these different microservices, all these different remote data and systems that you don’t control because today’s business is done through an ecosystem of suppliers and it’s more distributed than ever. You don’t do your own payroll, you outsource that, you don’t do your own all kinds of things in financial services transaction. The clearing is done by clearing a broker. There’s all kinds of different interactions throughout the economy, and so to bring all that data together so that the AI and the agenda, AI knows what it’s working on, data, fabric technology is how we make that simple. And so combining data fabric with the Appian process engine, the workflows, and then having an interface to that to the humans because you still want humans in control, humans setting the goals, humans supervising and orchestrating. That’s how it all comes together with the data fabric, making sure that you have the right data.
Isaac Sacolick:
John, want to go to you for a second before I go to the break? You got all these capabilities that are coming to fruition. What are some of the use cases that you think about?
John Patrick Luethe:
Well, I did an activity based costing exercise for the insurance companies and we’re looking at TV agents and it was a workforce of a couple thousand people and we looked at how they spent some time and we divided it into add value, doesn’t add value, and we looked at the stuff that obviously things that don’t add value, want to get rid of the things that add value. We looked at is it automated or is it not automated and how much you can do for the time savings. And when I see the capabilities come out to Stitch during the AI into business applications and workflows, and then I look at how much time people are doing really manual things, the opportunity is out there and that’s what really, really excites me. It’s just being able to go back and find something that a thousand people do on a daily basis and how do we not do that task because, not because it’s not required, it’s an absolute required step, but how can we automate that thing and make it go a whole lot faster?
Isaac Sacolick:
Thanks, John. Thanks everybody for joining this week’s Coffee with Digital Trailblazers episode 1 0 4. Today we’re talking about accelerating transformation, why AI is boring in 2025, talking about the intersection of document processing, RPA intelligent automation, low code machine learning, and now generative ai, and now bringing Ag Agentic AI to our organization. Folks, everybody, this week’s episode is brought to you by Appian processes to find your business, make them better. With Appian, the leading platform for process orchestration, automation, and intelligence, the Appian platform empowers leaders to design, automate, and optimize important processes from start to finish. With our industry leading platform and commitment to customer success, Appian is trusted by top organizations to drive transformational process change for over 25 years. Amazing, Mike, I love that picture of all of the founders at the conference up on stage, all working together for 25 years around this mission that you’re describing as boring.
I want to start my next question with Joe and then we’ll get back to you, Mike after this. Joe, we’re creating a paradox here, right? We’re saying AI is boring, but for the last year and a half, it’s the excitement that boards and CEOs and quite frankly a lot of technology companies and even those of us in media have been shining a light on all these exciting, sexy, amazing areas, whether it’s autonomous vehicles, whether it’s robotics, whether it’s AI breaching into generative ai, that’s what’s getting everybody excited and a little bit fear of falling behind. And we’re coming here today and saying, AI is boring in 2025. Take these amazing capabilities and plug it into the most pragmatic areas of your business to drive value and lean on the intersection of all these different capabilities that have been around for a while. Lean on them because generative is adding an extra flavor of capability to be able to bring workflow to an amazing set of productivity. Connect people, connect data. Joe, how do we excite leadership around the business opportunities when we’re saying these are boring areas for the business to invest in?
Joe Puglisi:
I think there are a couple of things, Isaac that come to mind. We’ve held out the promise of the single pane of glass for management to be able to know what’s going on in their organization sort of across the board. And despite lots of efforts of BI tools and integration tools without the ability to orchestrate the flow of information among disparate systems, rationalize, align, and truly present that information in a cohesive way, which I think AI is going to have the capability to really do, we haven’t been able to deliver on that promise. And so it’s boring in the sense that we’re talking about the same old thing, really understanding what’s happening in your business. But I believe for the first time with tools like Appian and other workflow and orchestration kinds of tools, we can really do it. We can go to all these different systems and understand the data and build that true perspective of what’s going on and even aid in the decision making about who, what, when and where to move the parts around.
Isaac Sacolick:
So Joey, you’re going to have to go deeper from me here, right? We’ve been selling this for a while. The CEO wants to be wowed, the board wants to be wowed, and we’re going to come back to them with a use case and a set of use cases that on the surface seems boring but have a tremendous opportunity. How do you sell that in? How does a digital trailblazer sell that in as a priority going into 2025?
Joe Puglisi:
I think you highlight the pitfalls that the previous attempts have fallen into. What have been the wrinkles misaligning of data, the lack of a true understanding of implications of certain decisions. AI has the ability like people to look at the numbers and say, well, wait a minute. There’s something a little askew here. Let’s figure this out. It can integrate more, it can integrate faster, it can have many, many more rules to follow. So I think you can paint a much broader picture of what you’re able to present and the quality of what you’ll be able to present to management. I truly believe in it,
Isaac Sacolick:
Mike, help our audience here, right? We’re really excited by this idea of document processing. I think that’s probably the area that every major company has struggled with over the last 10 years, and now we’re bringing all this gen AI capability to it. We’re bringing low code capability to it, but how do we sell this in so that the executive committee and the board see the value out of it and get over this, I need to put my eggs into the hike basket.
Mike Beckley:
Well, I think this is really easy. It’s budget times and when you look at your traditional OCR and it’s costing you $14 million a year and it’s got straight through processing rates that are still only in the high 60 percentage points, maybe 70 if you’re lucky, and you’re able to say, well, now generative AI directly, like Joe was saying, addresses the edge cases. It addresses the exceptions that we’ve been talking about. That means that we can dramatically cut our spend upfront on the software. It’s going to be less expensive to use generative AI to process that paper. And because there’s going to be far fewer errors, we’re going to get higher straight through processing. That’s a lot less manual labor and that’s more indirect savings that will translate into millions more. And so that’s the simple part, but ghost reporting right behind handling all that paper is the creation of new paper.
That single pane glass is really another way of saying people need to create dashboards and reports and spend the white color workers spend a whole lot of their time, maybe 20, 30% minimum on generating reports for their bosses, and generative AI is helping solve that problem. Providing ways to automate those data reporting pipelines and gather information more efficiently through data fabrics and attaching a data fabric to a generative AI engine in a secure way allows us to not have to know where the information lives, the generative AI can go find it for us and show us its chain of reasoning and how it got to create the queries, and then automating that pipeline so that we don’t have to go create a manual report every time we’re asked for one by the executives on how that new product is doing and how’s that product launch going and how’s the adoption go? All of that can be generated for us and let people go back to thinking and reasoning over what to do about the results rather than trying to find out what the truth is.
Isaac Sacolick:
Yeah, Michael, I think that’s a good way of showcasing that. The intention was always there. It was actually a lot harder for us to implement a lot of the things we were discussing, and then we got to the point where we had all the data integrated, we had all the capability there, and we fell short because of all the exceptions that you were highlighting at the beginning. There’s a lot of messiness in human decisions when it comes to complex processes, and that’s what we’re trying to bring to the table today. Go ahead, Joanne. You wanted to jump in on this?
Joanne Friedman:
Yeah, I did because in addition to cost savings and value creation on the value creation side, whether it’s agentic or just generative, and you can’t necessarily use generative AI for everything. It’s not a one size fits all, but in the cases where you are using it as part of angen capability, you have something different than a single pane of glass. You have the opportunity to create context around each of the panes in the single pane and go deeper, go broader, go wider. That gives you a different kind of perspective. And that perspective is really what seems to be resonating with the c-suite around it because if they’re purposeful about what they want to accomplish using the shiny new tool and it’s presented to them as an opportunity to broaden the way they make decisions or incorporate other factors into those decisions, then they’re out of the silo of the process of decision making. They not even think bigger and by thinking bigger, they get to more cost effective value creating decisions, so revenue, growth, innovation, resiliency, all those top line things now become doable. We’re not just looking at process optimization for costings, we’re leveraging the same data to be able to do those top line values just as much.
Isaac Sacolick:
Excellent. John, top line values, what do you want to bridge off of that on
John Patrick Luethe:
Going you? Just to respond back to the other one about the democratization of data, if I can, you can comment on that really quickly.
Isaac Sacolick:
Yeah, of course.
John Patrick Luethe:
And I think over the last, I’d say 10 years, the people have been a lot more willing to consolidate all the data to give in data lakes and the generative AI has been such amazing informational retrieval tool, but we’ve also had people using, I would say power BI and things like that. It’s been so nice to be able to actually treat the dashboards you want, but there’s a whole lot of people, they can’t create the dashboards because they don’t have the skills for it or they get a dashboard and it doesn’t quite have what they want. I think power generative AI is ability to write code or change things. I think is also going to just help really close lot of the information that people want to see on the dashboards. I think if somebody’s able to get the data consolidated into a data lake and somebody’s able to get a report and it doesn’t happen what they want, it’s going to be so much easier to say, have the generative AI tailor the report to what you need to get exactly the data you want or create a new report for things or query that you need for the data.
And so I think from surfacing data for executives for reports to get the information people need to make decisions, I see a lot really good stuff in that space on it.
Isaac Sacolick:
Yeah, I think John, to your point, I mean the fact that so many people on our staff in the employee base have jumped on being able to test and evaluate and use LLMs, that will be a bridging point for them to work with these other capabilities and maybe just lower the change management barrier that we’ve seen when we brought new capabilities into our platforms. Liz, welcome to the floor. How do we make a boring use case? Sound exciting? Oh, Liz, can
Liz Martinez:
You hear me? We can hear you hear me. Excellent.
Yeah, so first of all, I don’t think there’s anything boring about making money, so let’s just take that right off the table. The idea about these boring use cases allows us to take something that we know is a huge cost of doing business either operationally or potentially cost of good sold just in terms of the labor intensity and reduce that down so that we can then invest in things that make a much higher value add to the company strategy. Now, the hard part is actually doing the business case around that. Everybody likes to say how, oh, this is going to be so great. We’re going to set up these internal l lm, it’s going to be private lm, and we’re going to do all this great AI stuff and blah, blah, blah. That’s really expensive. I’m sorry. It’s really expensive. And so you actually have to do the hard work of doing the business case, building out those AI tools, those agents, whatever it’s that you’re going to do actually has to be offset with the value that you’re going to get so that you can, and maybe it’s an ROI over multiple years, or maybe it’s within a year, I don’t know, but putting pen to paper and demonstrating the value add in of dollars and when those dollars can be then redirected to something that’s more strategic for the company.
That’s how you get UR Csuite engaged.
Isaac Sacolick:
So start with the money, which is usually a place that we’re all fearful of and yet we have some pretty good interesting use cases. Mike, you brought up document processing earlier and Joanne brought how do we show growth and potentially long-term value out of this? What I love about document processing is being able to bring my corpus of documents, my 2, 3, 4 years worth of documents back into an environment that’s intelligent, and then being able to ask questions around it and saying, how do we get smarter as a business around this? Joe and I have seen this in the construction industry. How do we bring all of our bid documents, all of our planning documents into a single environment and ask a very simple question, what projects are we more profitable on? Which ones should we bid more aggressively on, and which ones do we have to get more efficient in our operations around before we start bid around? I really like this idea of now I have a way of bringing all this intelligence from 3, 4, 5, 7 years of documentation that we have and using that to some kind of competitive advantage. Joanne brought up healthcare. That could be a really exciting place for building efficiencies and also developing smarter and more personalized healthcare by looking through all that documentation that we have there,
Liz Martinez:
Michael, and that’d be part of the business case, right? That actually you estimate a percentage of your business that you can actually create some efficiencies on and you include that in the business case.
Isaac Sacolick:
Absolutely. Michael, tell us some more of these exciting use cases that come from the boring side of gen ai.
Mike Beckley:
Yeah, so let’s say the most boring one is you mentioned earlier about software development lifecycle. People have this dream and this vision that generative AI will suddenly replace all software development and on its way to achieve singularity and placing all humans, the first thing that’s going to do is write all the code for us. And so we don’t need to worry about it. I don’t mean to be negative, but that doesn’t matter. It’s not going to happen in 2025. Don’t put it in your forecast or your savings plan unless you really want to look silly. What is happening though is the most text heavy part of the software development life cycles requirements, and so we can automate the heck out of requirements, and that’s where it works today. We can ingest all those requirements documents for the most complex applications and create very detailed plans for how to maximize the reuse of existing components and systems and data integrations and minimize the redundancy and the cost and the risk of building and modernizing your ERP systems.
So this is what everyone’s doing. Everyone’s trying to modernize their systems so they can take advantage of ai, but AI itself, where it’s doing it is not automatically replacing all software developers. It is automating the most text heavy part of the SVLC, and that is upfront requirements management and planning for these application build. And then so that’s another great boring use case for you. But I do want to emphasize when I talked about pushing paper and OCR in these business cases, I’m saying it for really specific reasons. I don’t think you need to get esoteric with the future value when you’re talking to executives because they get jaded quickly. They’ve been promised and over promise what new technology is going to deliver in cost savings. But when you start very specific on OCR, you can prove it in days and weeks, and that is what people need.
They need to see proof and real value from generative AI applications, and that is one way we can sit down and run away, feed in the documents like you were talking about Isaac with construction documents, planning documents, whatever they are underwriting documents and show executives absolute knock down amazing results virtually overnight. And so that’s what’s going to get funded. That’s what’s going to be effective. And yes, of course that will have value in terms of better customer experience. They’re getting their insurance quote in seconds and not in weeks. They’re better employee engagement because they don’t have to sit around doing all this manual, be keying and twin systems, but don’t promise that. Just promise what you can show and demonstrate, which is you can scan a whole lot of different paper than you ever could at much higher quality than you ever could before, and that leads the foundation for transforming all of these human workflows. And so no, it’s not as sexy as exciting as the singularity. It’s not as cool as worrying about whether or not generative AI is going to cause nuclear war, but why will it be a massive improvement in your bottom line?
Isaac Sacolick:
Mike, you’re highlighting a piece that I think is incredibly valuable for anybody who’s worked in the low-code and no-code space, the ability to take a concept and show results pretty quickly that we’re heading down the direct track, that we’re demonstrating value around it. I mean, I think going back to the question I put Joe on the spot, how do you get the board excited over this? Tell me if I’m sitting on top of all this information, if I’m all the underwriting data that we have access to all the tax documentation that the government has, what can we do with it quickly and start showing and getting people excited about this use case of looking through this documentation? What can I do in a short amount of time?
Mike Beckley:
Well, whatever you’re doing with it today, what you’re doing with today is leaving most of it behind, but you’re automating some of it with maybe 70% breakthrough processing investing. And so to be able to take that data and now feed it through generative AI and invent AI powered pipeline into your workflows that you already have, don’t create, invent a new workflow that takes too much time. Use your existing workflow and pilot using new generative AI techniques, and you can overnight improve this rates. You’re processing 20, 30%. You can actually get much better accuracy on more fields, on more data types, on more document types and show those results right away. That’s what you do. Don’t invent the future. Just reinvent what you have right in front of you. With this new technology.
Isaac Sacolick:
Saji has been trying to raise his hand. He asked me a question over a message. We start bringing all these capability and we start exposing it to our employees. How can we then get creativity using AI now that we have access to all this information? Mike, I don’t dunno if you want to take that or if somebody else wants to take that, but I think it’s an interesting question. We keep bringing more intelligence and more capability, and now we’re doing it inside people’s workflows and we’re saying, Hey, we’re picking away the difficult work that you were doing before and we’re bringing more creativity capabilities to you, Joe, what are some of the creativity that we can bring to them? Thank you, Joe.
Joe Puglisi:
Look, nobody likes to open a spreadsheet, copy a couple of columns, open another spreadsheet and paste them. Then go to the ERP run report, seven dash a copy, a few numbers off there, but this work is mind blowingly boring, and if we can introduce tools, low-code tools like Appian or other tools that we can teach our employees how to do their functions, just what they’re doing today, just do it faster, smarter, and with a higher degree of quality and free up their time, they’re going to embrace that. I’ve long been an advocate. You see me post all the time about how corporate America needs to invest in its existing employees, and this is one of the best ways that you can take your current works out and elevate them and give them the ability to do the mundane work, hand that off to an agent, teach ’em how to hand that stuff off to an agent and let ’em add more value in ways that AI isn’t capable of yet.
Isaac Sacolick:
Go ahead, John.
John Patrick Luethe:
I was at Stanford for 10 years and I remember being a new analyst and Joseph gave my be nightmares work I used to do back in the days. Yeah, yeah. So that totally resonates with me. The other thing I was going to say is that I have seen that the jury’s out on how much generated AI helps on development, productivity, know pretty large company, they went all in on it talking a year later on, how much of efficiency gains did you get? And I think you guys study that more. I do know that anything that you code you do get from generative ai, man, the amount of testing you have to do, it goes massively up. And so that’s an area I think that it’s, you almost have to look at how much you have to increase in testing for whenever you bring in this third party technology or third party stuff, and with generated ai, it can give you different results at different times, indeterminate nature of it. So yeah, I remember being so much, so much busy body work than I said just starting off in my career. It was hours of the days of it.
Isaac Sacolick:
Look, I subscribe and I’ll go to Joanne. Joanne, you remember the days when mobile first came out and cloud first came out and we kind of went after the low hanging fruit. What does it help us optimize that we haven’t done before? But the real exciting part of mobile was when we built mobile first interfaces and extended people were doing from things that were doing in the office to things they were doing out of the office, that became a whole new set of capabilities. Same thing with cloud. What it enabled us to do is scale things that we couldn’t do before and get access to capabilities that we couldn’t do before. I think when you start putting together RPA intelligent automation low code, you put all this together document processing, and you start bringing into the workflow people and say, start using this and start thinking differently about what you’re doing. Get off of trying to do copy paste spreadsheets and start asking this thing questions. We don’t know exactly what people are going to start using this for. We want them to work with us and partner with us to figure that out. Go ahead, Joan.
Joanne Friedman:
Well, we definitely want them to work with us and partner with us, but we’re freeing up the workforce to be innovative, to experiment, to have the time to think that they didn’t have before. But really, if you look at it from a data fabric perspective or a data perspective, there are common data elements across different workflows or different process streams, and this is where agen AI really comes in because once you have those common denominators and you start adding different contexts, I like to think of it like a diamond, right? When you look at the diamond and you look at the light hitting the stone in one way and then turn the stone a different way, you get a completely different perspective and it makes you think about things in a way that starts adding value. I mean, the creativity that one can use with generative AI is one thing.
When you start combining the mls, the machine learning and specialized process controls, if you’re in manufacturing or you’re making things and all the different kinds of other kinds of ai, that’s when the world changes because the mundane tasks that you started with ends up being, wow, I just took all that paperwork, got rid of it out of my sort of day. It is all done. Now I can think about how can I make a better product? How can I take people’s feedback on a product, whether it’s an insurance policy or a cup, and say, how do I make this better, faster, cheaper, more enticing to my customer, more enticing to my customer’s customer? That’s where we’re giving knowledge workers actually the ability to become knowledgeable because we bring in all the different streams of data in new ways. It’s kind of like Lego, right? If you only work with white blocks, all you’ve got is something with white blocks. If you start mixing and matching those and make them light, microservice light or containerized light, you can build whatever you want, and that is, I think, the greatest value of this inflection point. As AI is getting more sophisticated, we have the ability to create
Isaac Sacolick:
Mike to tee up for you. What are we creating? You were going to chime in on this.
Mike Beckley:
Yeah. Well, what I was going to say was how do you spark that and scale it that creativity, and the way we’re doing that with our clients today is we’re able to run hackathons where business leaders, business users, 50, 60 of ’em at a time, who have not previously built Appian applications. They’re not experienced low code developers. You give ’em a two hour enablement session and then they game out with workflows that they already deal with. They want to automate and using low code generative AI and agent building techniques on the existing workflow in a single day. They can innovate and create, but what comes out of that is on platform, which is already designed to scale and be governed by it and is already regulatory approved and compliance, so they can not just have creativity, but that creativity can then actually be put into the most direct way in operations and workflows and scale that across an enterprise. And that is the real magic of what we’re talking about here by saying, look at the workflows you have, use these low-code, generative AI agent technologies in combination with your existing workflows, and that’s where people can all be part of the process of change. People can all be involved in the process of operationalizing how AI is going to actually transform the most boring parts of the work.
Isaac Sacolick:
Mike, both John and I had the same question. What types of applications are coming out of these hackathons that your business users are able to spark out that quickly?
Mike Beckley:
Yeah, so it’s operations. It is the middle office and back office operations of how do you actually perform a treasury financing operation? How do you move money from one account to another? How do you actually onboard a customer more efficiently when you have to reconcile many different systems to accomplish that today? How do you move from just onboarding a customer to then updating the regulatory reporting for that customer? It is in the operations groups where you have the most pent up demand to try and tackle those customer and retail operations. I think that is the immediate benefit from holding those types of hackathons. As long as it’s not just about the ai, it’s about how do you look that through a workflow perspective.
Isaac Sacolick:
Mike, I’m going to let you end with that statement. I think that’s incredible advice to all the digital trailblazers listening here in terms of bringing boring use cases. Bring the capability to your operations teams. They know what the steps in their process are. They know where they’re struggling with too much work in the boring areas and give them the ability to do a hackathon, bring the capability to them where they can experiment with these technologies. The fascinating thing about this now is we have data fabrics to bring data in. We have low-code capabilities to enable building out the workflows. We have document processing to load in all this documentation. We have all these different capabilities, and the reality is, let’s bring it to a workflow. Let’s set up our agents where people are actually working and let’s let them feel empowered. Mike, any last words for the group today before we close out?
Mike Beckley:
I think you’ve closed out. Well, thank you, Isaac, and thank you to the panelists. I think it’s been a great conversation. I know we’re going to have incredibly boring in 2025 by the time of real actual practical value from Generat ai.
Isaac Sacolick:
Thank you, Mike, and thank you, Joanne, Joe, Liz, John for joining me today in this discussion on how AI will be boring in 2025. I want to thank our sponsor today, Appian Business and Organizations run on processes, make your process better with Appian, the process company. Visit appian.com to learn more. Thanks again, Mike, and to the Appian team for sponsoring today. I just want to let you know about our upcoming episodes next week, holiday week. We’re going to skip November 29th. Here are your three episodes for December. I just announced this yesterday, December 6th. We’ll be doing culture transformation, evolving, diversity, inclusion, hybrid working and global collaboration. A lot of changes I expect happening in companies over the next few years, so we’ll be talking about that on December 6th. On December 13th, AI area transformation, gen AI guardrails, and how to implement safe gen ai, and then on the 20th shaping tomorrow moral courage on taking the right path and doing the right thing. All three of those we’re recommended by listeners. So if you’re listening here and would like to share an idea for a topic to message me, again, thank you, Mike, and thank you Appian. Thank you, Joanne, Joe, Liz, and John for joining me today. Happy Thanksgiving to all of us who are celebrated here in the United States and everybody have a safe and happy weekend. Have a good one.
Participants Isaac Sacolick Joanne Friedman Heather May Joseph Puglisi Martin Davis John Patrick Luethe Episode Summary AI is increasingly automating and transforming traditional IT...
Isaac Sacolick hosted a "Coffee with Digital Trailblazers" episode focused on AI in nonprofits ahead of Giving Tuesday. Nonprofit leaders shared their experiences and...
Participants Isaac Sacolick Joanne Friedman Heather May Joseph Puglisi Martin Davis Derrick Butts John Patrick Luethe Liz Martinez Niraj Tenary Episode Summary This week’s...
