Sandy Kemsley’s Vlog - AI and BPM
Sandy Kemsley Photo

AI and BPM

By Sandy Kemsley

Video Time: 8 Minutes

Hi, I’m Sandy Kemsley of I’m here for the Trisotech blog to talk about the latest Hot Topic in BPM: Artificial Intelligence.

Now, I’ve been at a couple of conferences in the last month, and I’ve had a few briefings with vendors and there’s a lot of interest in this intersection between AI and BPM. But what does that mean? What exactly is the intersection? And it’s not just one answer, because there’s several places where AI and Process Management come together.

Now, the dream, or the nightmare in some people’s opinion, is that AI just takes over processes it figures out what the process should be then it automates and executes all the steps in the process. The reality is both less and more than that. So, let’s look at the different use cases for AI in the context of BPM. Let’s start at the beginning with process Discovery and design and there’s quite a bit that AI can do in this area as an assist of Technology. Now, at this point it might not be possible to have an AI completely designed processes without human intervention, but it is possible to have AI act as sort of a co-pilot for authoring process models or finding improvements to them.

There’s a couple different scenarios for this.

First of all, you could have a person just describe the process that they want to have in broad terms, and have generative AI create a first impression of that, or a first version of that process model for them. So, then the human designer can make changes directly or add additional information, the AI could then make refinements to the process model and so on. Now, the challenge with using generative AI in this scenario is that you need to ensure that the training data is relevant to your situation. This might mean that you need to use private AI engines and data sources that are trained on your own internal data or on data that’s specific to your industry in the very least in order to ensure some reasonably good results.

Now, the second process modeling scenario is when there are logs of processes in place, like we would use for process mining, and we’ve talked about process mining in in a previous podcasts. Now, in that case, there are possibilities for having AI look at the log data and then other enterprise and domain data and using process mining using other search-based optimization suggest improvements to the process. So, for example adding parallelism at certain points, or automating certain steps or decisions, or having some activities be required for regulatory or conformance reasons. Again, there needs to be some checks and balances on the training data that’s used for the AI to ensure that you’ve included processes and regulations that pertain to your business.

Now, in both of these cases, there’s the expectation that a person who’s responsible for the overall process operation, like the process owner, might review the models that are created or revised by the AI before they’re put into production. It’s not just an automated thing where the AI creates a model or modifies a model and it’s off and running. Now, we can look at using similar types of a AI and algorithms that you would for process improvement that are based on process mining and other domain knowledge, we can also use those in the scenario where AI acts again as a co-pilot, but for the people that are doing the human activities in a process, so the knowledge workers. Now they can ask complex questions about the case that they’re working on, they can be offered suggestions on the next best action, they can have guard rails put in place so that they don’t make decisions at a particular activity that would violate regulations or policies.

Now, we already see a lot of decision management and machine learning applied in exactly this situation. So, a knowledge worker just needs a little bit of an assist to help them make complex decisions or perform more complex activities. And adding AI to the mix means that we can have even more complex automation and decision-making that can support knowledge workers as they do their job. So, the ultimate goal is to ensure that the knowledge workers are making the best possible decisions at activities within processes, even if the environment is changing maybe regulations are changing, or procedures are changing. And then also to support less skilled knowledge workers so that they can become more familiar with the procedures that are required because they have a trusted expert, namely the AI, by their side coaching them on what they should be doing next.

Now, the last scenario for AI in the context of processes, is to have a completely automated system or even just completely automated activities within a process that used to be performed by a person. So the more times that an activity is performed successfully, there’s data collected about the context the domain knowledge that all that go behind that decision, the more likely it is that AI can be trained to make decisions and do activities of the same complexity and with the same level of quality as a human operator. We also see this with AI chatbots. We’re seeing these a lot now, that where they interact with other parties processes like providing customer service information. Now, previous previously a knowledge worker might have interacted with a customer maybe on a phone or by email, we’re seeing a lot of chatbots in place now for customer service scenarios. Now, a lot of them are pretty simple they don’t really deserve to be called AI. They’re just looking for simple words and providing some stock answers but what generative AI is starting to give us in this scenario, is the ability to respond to more complex questions from a customer and leave the human operators free to handle situations that can’t be automated or rather can’t be automated yet.

Now, currently I don’t think we need to worry about AI completely taking over our business processes. There’s lots of places where AI can act as a co-pilot to assist designers and knowledge workers to do the best job possible. But it doesn’t replace their roles: it just gives them an assist. Now, a lot of Industries don’t have all the skilled people that they need in both of these areas for designers, for knowledge workers or it takes a long time to train them so letting the people who are there be more productive, is a good thing. So, using AI to make the few skilled resources we have more productive is something that’s beneficial to the industry it’s beneficial to customers. Now, as I noted earlier, the ability of AI to make these kinds of quality decisions and perform the types of actions that are currently being done by people, it’s going to be heavily reliant on the training data that’s used for the AI. So, you can’t just use the public chat, like chat GPT, for interacting with your customers. That’s not going to work out all that well. Instead, you do want to be training on some of your own internal data as well as some industry specific data.

Now, where we do start to see people being replaced is where AI, is used to fully automate specific activity, specific decisions, customer interactions within a process. However this is not a new phenomenon. Process automation has been replacing people doing repetitive activities for a long time. So, all that we’re doing by adding AI, is increasing the complexity of the type of activity that can be fully automated. The idea that we’re automating some activities is not new, this has been going on a long time. So, the bar has been creeping up: we went from simple automation to more complex decision management, machine learning and now, we have full AI in its current manifestation. So, we just need to get used to that idea that it’s another step in the spectrum of things that we’re doing by adding intelligence into our business processes.

Now, are you worried about your own job? You could go and check out or just look around at what’s happening in your industry. If you’re adding value through skills and knowledge that you have personally that’s very difficult to replicate, you’re probably going to be able to stay ahead of the curve and you’ll just get a nice new AI assistant who’s going to help you out. If you’re doing the same thing over and over again however, you should probably be planning for when AI gets smart enough to do your job as well as you do.

That’s all for today. You can find more of my writing and videos on the Trisotech blog or on my own blog at See you next time.

Blog Articles

Sandy Kemsley

View all

All Blog Articles

Read our experts’ll blog

View all

Learn how it works

Request Demo

Confirm your budget

Request Pricing

Discuss your project

Request Meeting