Continuing his conversation with Ian Johnston of AI4Process, Alexandre Nevski explores the artificial intelligence capabilities of Pega. Together, they cover its use cases and the right approaches to mitigating its potential risks. Ian explains why considering the people aspect of digital transformation must never be set aside, emphasizing the need for truly capable individuals to carry out your implementations. Ian also talks about his secret to reuse and why he thinks process AI would probably get the biggest business benefits in the long run.
Pega® is a trademark of Pegasystems Inc. Please visit https://pega.com to learn more.
—
Listen to the podcast here
Pega AI, An Extended Conversation With Ian Johnston
Introduction
Last time, we covered the iterative approach, funding models, and the importance of people in digital transformation. However, with many years of experience in the field, Ian’s insights were too numerous to fit into a single episode. As the Cofounder of ai4process, Ian is ideally placed to tell us about artificial intelligence. In this episode, we cover the latest use cases, consider risks, and discuss the AI capabilities available in Pega. As a bonus, you want to stick to the end of the episode as he reveals his secret to reuse. Let’s start with Ian’s description of his professional journey, only a fraction of which made it into the first episode. Without further ado, let’s continue the conversation with Ian Johnston.
It’s an immense pleasure to have you on the show. I’m interviewing a lot of Pega experts for this first season, but you’re by far the one that has been involved with this technology the longest, isn’t it? I was wondering maybe for the readers, if could you share how you got into Pega, I think 30 years ago now. Also, what made you stay in that ecosystem all this time?
I started off life in banking in the city, in trading systems and banking systems. I did a project for HP for the company I was working for at the time. They had this workflow system, which I thought was cool, where you could build a process, controller process, and user interface. At that time, I was headhunted by a guy who was working on behalf of Pega and trying to build up the Pega team in the UK. I already had an interest in the workflow. He was a company that was starting off in the UK looking for people like me. It sounded like an interesting subject to get into.
Was the technology already on your radar, or is it something that you’d like to get more experience with? Once you got into Pega, what did you find?
That was in 1994 when I started with Pega. It was the mainframe Vaxx VMS-based predecessor to what Pega now has, Pega Infinity, which was the old Pega Works system. They had a small number of clients in Europe, building a team and trying to build things up in the UK. We started off as what we call client consultants. We did everything from being the account manager to designing, building, and managing the team on a project. It was a pretty full-on role.
Is that part of what made you stick with it at the beginning?
The great thing about what Pega did with the product over the years was they were continually innovating. They were one of the first people to start using a markup language, which turned into HTML, for instance, for the UI, and all sorts of other areas. They kept evolving the technology and the breadth of the functionality. Your learning curve wasn’t taking a dive. It was always interesting, and you are always up to date in terms of technology and where it is going. It was kept interesting, and that’s the main thing. I’m a techie at heart. Obviously, you have to perform many roles as you move through life as a consultant, but from a technology point of view, it always kept the interest up.
The evergreen aspect of Pega technology is that you go through these waves of modernization, and you’re not necessarily doing the modernization, but you are taking on the wave yourself while accompanying your clients. In my case, I remember at least two big transitions between different UI technologies, UI stacks, and we would take those clients from one generation and one set of technologies to the next without necessarily having to throw away everything. That’s what you were talking about.
There was a natural progression for clients of Pegas to be able to move from one technology to the other, probably not moving from works to what Pega rules were at the time, but now Pega Infinity. Moving from one to the other was not something that you could do to upgrade and run some code and do it. You probably wouldn’t want to either because typically, when you have a technology that’s 5 to 10 years old and you’re moving to the latest one, what you do and what your business does and the processes and what the capability of the tool is so much better that you don’t want to do a straight upgrade anyway. You want to do a redesign and a rebuild.
What you do have is because you’ve already got a system, you’ve got a great set of basic requirements that you can then look at and say, “How am I going to make this better? What do I need now? What’s the technology capable of?” Upgrading from 1.0 to 2.0 is not necessarily what you want to do.
It’s not a straight upgrade. The people stay as well. You’re upgrading the technology, you’re keeping some things, and like you say, you already have your specifications, but then a lot of the people that have invested time into learning the technology can continue to evolve with it.
It’s something like Pega and a lot of other tools, essentially, they stay the same and their core stays the same. Your expertise in moving from Pega works as we had originally to Pega rules was the same people who had all the knowledge on works were then able to very quickly move into the new environment and be fairly quickly efficient and knowledgeable about that technology as well because the fundamentals are not changing. There are other things that change around the edge.
–
Blueprints Of Innovation
Ian, for that second segment, we usually focus on tools and technologies, and when we were preparing the interview, we mentioned artificial intelligence. It’s everywhere these days. With your amazing and very wide experience of Pega, how do you think this is going to change that ecosystem?
There’s a lot of stuff that’s evolving, and there are a lot of ways people are potentially looking at using AI in their businesses. There’s a lot of hype and talk around it. Coming up with a use case in your business is if you’re not using something that everybody else is doing off the shelf, then it can be quite difficult. You need to have bought into the art of what you can do within the organization. You need to have looked at that and then you need to know what you can do and how it would benefit you in order to get the best out of it using a couple of examples for instance. You can use our AI in a process to determine which route you take down a process based on what happened before, what was the most effective way of handling this particular customer issue or whatever.
You can use AI and machine learning to get better at it and things like that. That’s one way, but you need to have worked out at what point in your business processes this is applicable and how much benefit is it going to provide. You need to be quite intelligent and in terms of understanding how it can benefit and what it’s capable of doing. The other type of AI that’s quite interesting at the moment is more the generative stuff, which can help you write an email or a book and design an application by writing a few words, which is great. You can use it to do something small like writing an email to a customer or you could, like the Pega blueprint, ask it to build you an application and it will build you an initial application design that you can then put into your Pega system and it will create something which will be a starting point.
You need to be quite intelligent in determining how your business processes can benefit your target audience and what they are capable of doing.
It’s an accelerator to get where you want to go, and that’s great. I think the process of AI, where you determine how to handle real-life situations with customers or whatever, is probably the one that is going to get the biggest business benefit in the long run. The generative stuff will help you get from nowhere to something to start with very quickly, which is good, but applications are seldom that easy. It’s very easy to create demo systems. It’s very easy to create something basic and something as a starting point, which is great, but it’s never going to be where you want to end up.
You’ve always got to take that, enhance it, build on it, change it, and get to where you’re going to go. Out of a project life cycle, how much of the project is that piece that initially creates something that is demonstrable versus how much of the project is then to get that into production, it’s quite a big proportion of the work you’re going to do.
Those are the two areas that seem to be prevalent in our world at the moment, the process AI and then the generative. It’s up to us in this industry to point out where the quick wins are, where businesses can take something that’s out of the box in the way the application works and can be immediately applied and a lot of businesses would benefit from that kind of thing. Getting into the core business and integrating machine learning into the processes is something that requires a lot more thought. Intelligent marketing has been around for quite a long time. It’s the newer stuff, which is, “How do I utilize AI in my processes? How do I utilize AI in the generation of artifacts?” is the newer stuff.
Focusing for now on the process part, what kind of approaches have you experimented with, especially which ones work in order to identify these use cases with clients? Are there any lessons learned already that help organizations narrow down this part of discovering the art of the possible?
It’s having done something previously with a customer that you can then replicate with a client, rather, you can then replicate with further clients that are in a similar industry. For instance, in ai4process, we have a collections framework. We’ve enhanced the collections framework from Pega and we use that in a number of our clients, and we use the AI capabilities within that to essentially improve collection rates with customers when you are going to predict which course of action is going to give you the best returns.
That’s an established way of doing it. Most businesses that sell things to customers have collections, departments and need collection systems and therefore, it’s something that could be applicable to any business. To go into a business from scratch and try and work out where they could use AI and machine learning in their processes is something that needs a lot of expertise from whoever is going in to do the study. It requires, again, expertise in the business to help work out where they could gain the most from this type of technology. I’m sure most businesses can gain a lot in one or other aspects of this technology, be it decision-making, generative, marketing, or whatever.
I’m sure all businesses can, but it’s coming up with those use cases that are going to give them the best benefit based on the cost and on the fact that you need some pretty intelligent people in your organization to help you from organizations like ours to get there. Once they’re in and up and running, they require constant attention and maintenance. Generally, you need to monitor them. You need the models and the use of the models within the organization. It’s not something that you are as likely to do as in a traditional model where you create something and you put it live and then it runs until it’s superseded, enhanced or whatever. This is going to need attention all the time and your data is going to need constant maintenance.
What changes from an organizational perspective? Are there more operational roles because of that?
Clearly, yes. You’ve got your data scientist roles, but you’ve also got a wider role in terms of maintaining the whole AI capabilities across your business processes and making sure that they’re doing what you need them to do, monitoring them, tweaking them, changing the algorithms, changing the models, whatever where necessary. I think that’s an ongoing role.
You always have to look at where the money comes from to get more return from providing business processes to a particular area and fund your wider digital transformation.
I should have probably started with that, given how new all of this is. I’m pretty sure that some people in our audience will not know too much about the machine learning available for the process side of things. Maybe we should briefly summarize what we’re talking about. Would you mind?
Sure. I suppose we touched on a little bit earlier where you may have a business process, which, in its simplest form, you could take route A or route B. What machine learning would allow you to do is to look at what you know about this particular situation, be it a customer situation, a transaction or whatever. What you know means that you have a better chance of success if you go on route A or B, then what you need is a feedback loop.
If you send a customer down route A, for instance, and it doesn’t succeed, that gets fed back into the model so that next time, he’s more likely to take route B for instance. It’s thinking about where in your organization’s business processes you can benefit from the system learning which treatment of which customer is going to yield you better return based on everything you know about that customer and all the customers that have gone before.
This is where you need quite a bit of data for these models to be able to work.
It depends on the frequency that these decisions are being taken on. If it’s high frequency and you’re getting very quick feedback, then it’s not going to take very long in terms of elapsed time for it to learn. If it’s a longer time period, then it could take quite a while to get the data to learn what is going to be the best benefit.
Coming back to the iterative approach that we discussed in the previous segment, it sounds like because of that data requirement would be somewhere further down the line, not something that you start with or at least when it comes to the process of AI. Is that a fair statement?
If you get a good buyer and a good seller who understands what they are implementing and selling, you will be successful.
How far down the line is dependent and there are prebuilt models as well for some situations that would already get you to somewhere, a starting point where it would be intelligent enough not to make the wrong decisions from the beginning.
Coming back to now the second part, the generative models, you were saying that it’s not going to get you 100%. It accelerates the way that you may start a new project with your application. What about from an operational side? What use cases do you see for that capability for operational use cases?
By operational, we are talking project lifecycle.
Not as tools for developers or other engineers but as tools for the end users.
Something as simple as writing a document, an email, a letter, or something like that would be an example for instance. That’s great as long as the generative algorithms are using data which your business is going to be happy creating this content with. One of the big issues that would stop businesses from using something generative is if the data is sourced from outside of the organization and therefore, you don’t know or you don’t have control of what gets produced. There are all sorts of implications there.
Including intellectual property concerns.
If the data is sourced from internal data sources, then within the business or within a data source, which is well-known or specific to what you need, then you can be happier that what it’s going to generate is not going to cause litigation at some point in the future or not do something that’s plain wrong and not right, which could cause your business some damage.
I was thinking probably one thing that could happen is that the typical shape form of our workflows is going to start integrating these steps like previously, typically, we’d have somebody does a piece of work and somebody else does a check for us later to make sure that everything was done correctly. These building blocks tend to be quite everywhere. Would you also expect that these generic building blocks would start incorporating generative AI phase and then give it to a user before the user does anything or before a human being does anything that there would be already almost for every step in the process, some work being done by a generative AI model?
Potentially, yes.
What are the limits where that doesn’t work, and where can we not do this? You’ve already identified one area, which is if the data is sourced from the outside, and if we are not in control of the data, what other risks are associated with generative AI?
You’re risking your sources. Let’s say you’re risking the things we talked about, which is the inaccuracy of data, copyright, or whatever.
How about losing control of the models, lacking transparency in some industries where it’s not acceptable, or that sort of thing?
Lack of transparency is a big one because, in a lot of cases, businesses need to be able to explain their decisions or how the decisions were come to and if you don’t have control of that decision-making process, if it’s outside the organization or even if it’s inside the organization, if you can’t explain it, that can be difficult. You could create something that somebody could perceive as sexist, racist or something of that order that the model has come up with, which a human being necessarily wouldn’t do because they have certain controls.
Businesses must be able to explain how they come up with decisions. If leaders don’t have control over their decision-making process, they will have a difficult time.
I can see that there are all sorts of things that you need to think about if you are creating something using generative that you’re then going to publish in any way, shape, or form outside or even inside of the organization as well because it will still come back to bite you. A lot of those risks are going to put a lot of people off in some businesses from taking the generative stuff on, and the process AI as well to a certain extent because you need to be able to explain how it came up with its decision in some circumstances. A lot of the AI models are able to do that, but some of them aren’t. You have to make that choice.
Beyond the Pega ecosystem, what other big changes do you see on the horizon as regards to artificial intelligence?
It’s the progress that everything is going to make and certainly, software is going to make. Up until this point, mainly, you’ve got humans creating software that does things better than humans can. Now, you’ve got the extra dimension, which is that humans have now created the software and now the software creates the software, and it can iterate and improve it much of a faster rate. You are going to get, “I would’ve thought the software is starting to write the software a lot more.”
Potentially, it’s going to be exponential with an improvement in what the software can do because now you’ve got the software writing the software. It’s so much faster at iterating than we are. I think that’s the big difference. It looks as though it’s going to take off fast, but there are going to be some businesses that benefit from it, and there’s going to be a lot of other businesses that are going to be waiting around to see what happens and don’t have the expertise, vision, or whatever to be able to take advantage of it.
Artificial Intelligence: Pega’s learning curve wasn’t taking a dive. It was always interesting and up-to-date about where the technology is going.
Innovator’s Playbook
As we are talking about the human side of digital transformation on the show, what can we recommend to the leaders in order to be ready for this revolution?
It’s about keeping up with what is practical at this point and the art of what is practically possible at the moment. Saying we want to use AI in our organization in some way, shape, or form is a good start, but you need to take a look at what it can do and what the potential is within your organization within the practicalities of what we have at the moment. Certainly, taking advantage of some of the more off-the-shelf capabilities would be a reasonable starting point.
Do you have a favorite example of maybe a slightly unusual application or use case for AI that you could share with us?
Probably not unusual. I think we’ve been playing a lot within our innovation with chatbots, for instance, using other technologies to Pega such as Camunda and things like that, such as using more generic AI models that are available. The thing that I’ve noticed is that not all but a few chatbots that you interact with online are doing stuff that is useful now, and they’re actually able to communicate with you much better than they used to. The lines between the human and the machine are blurred to a certain extent.
To me, the advancements in chatbots over the last years from things that were useless kept popping up on the website every time and obscuring your view of what you wanted to see to now when you want to do something with the customer service department, that they get you in much more quickly than picking up the phone and they’re able to solve your issues. Not in every case, there are still a lot of bad ones out there, but I’ve been pleasantly surprised by how good some of these things are and how human-like they are without being too naff or condescending or rude.
It’s amazing to think that we expected chatbots before generative AI to be useful. Now, in hindsight, it seems absurd that the way we would build those V1, version one, chatbots with trees of decisions was very clunky. What you’re referring to is that now we have the tools to make that kind of instant interaction much more useful.
The generative is a big part of that because they’re not just going down a decision tree, as you mentioned, and they’re able to add nuance and understand what you tell them rather than just looking for keywords in your sentence that you’ve put in there.
It’s both creepy and amazing, I think.
One of the most impressive is the fact that now, in the medical world for instance, we care for older people or whatever. They can give older people companions who interact with them who may be AI and they’re not real person, but they’re better than real people in a number of ways. They’re more patient. They’ve much better listeners. They don’t get emotional or upset or whatever. They’re available 24 hours a day. There’s a lot to be said for using this technology in places that are appropriate.
Definitely, as a compliment, I suppose. Not necessarily as a replacement, but it certainly can be something that brings additional connectivity or a feeling of being connected to people. I, for one, definitely enjoy that part of the generative AI tools. As we are about to wrap up, are there any final topics you’d like to cover?
There’s one thing we haven’t touched on, which is a subject that’s quite close to my heart, which is reuse. There are a couple of points here that I’ve found throughout my career that some people concentrate or think reuse is being able to create reusable code. A lot of the facts are missed that you can reuse ideas and get reuse out of not making the same mistakes again, which is another one. If you build something and you roll it out into production and it doesn’t work, that’s as good reuse example as actually building something that works well, and you roll out and they can then reuse it, so you know what not to build next time. That’s very important. That’s one example. With these new technologies, I think in reuse, you can build things very quickly, build functionality and business processes very quickly.
If you roll something into production and it does not work, it is a good reuse example. It tells you what you can reuse and what not to build next time.
You don’t need to reuse the code because you can rewrite it or rebuild it very quickly, but in the new way, that was slightly different from the old way, which didn’t work as well as you’d hoped. Now you can throw it away and rebuild it from a code point of view, but you’re not throwing away the ideas or the previous experience of not quite getting it right. If you think about a project lifecycle back to the lifecycle again, how much time from the beginning to the end of the project is stuff being built? It’s probably less than a quarter.
Everything else you do is a much bigger proportion of the project. If you have to rebuild something, but you already have a basic design, which you’re going to tweak the requirements, which are slightly different, it’s much quicker to rebuild something rather than try and reuse something that wasn’t 100% right in the first place.
Reuse is almost like learning at an organizational and team level.
You can spend a lot of time trying to build reusable code, making it very complex and difficult to maintain. You end up rewriting it anyway because it’s too difficult to understand how it works and maintain it.
By the time you’re trying to reuse it for the 2nd, 3rd, or 4th time, you’ve moved on with other ideas. Your initial code needs adjusting. That’s been good advice to finish on. Thank you very much for coming to the show. It’s been an absolute pleasure.
Me too. Thanks.
—
Episode Wrap-Up
That concludes our deep dive with Ian Johnston. In this episode, we covered the use cases and risks of Pega’s AI capabilities. I hope that provided valuable guidance to anyone looking to integrate artificial intelligence into their business processes. I feel we were lucky to have someone with Ian’s background on the show. If that made you think of another expert who would be a great guest, please visit our website and fill out the guest recommendation form. We have more topics and guest appearances lined up. Stay tuned for more Tales of Innovation that inspire, challenge, and transform. Until next time, peace.
Important links
About Ian Johnston
Ian started his career in the City of London implementing financial dealing room systems and wholesale banking systems. He then joined Pegasystems in 1994 and worked on many workflow and BPM implementations across financial and other industries. In 2019 Ian co-founded ai4process together with Amit Agrawal to help clients get the best from their digital transformation. They have brought together many experts in Pega and other tools to help them grow the business and provide resources and best practice to their clients.
Enjoy New Shows
More from this show
Explore
© All Copyright 2024 by Glaukos Digital OÜ