Optimizing Value Streams and Processes with AI at SKILup Days 2024

Hi everybody. Welcome to Skill Up Day, and thank you for joining me for optimizing Value Streams and Processes. I'm Donna Knapp.

I am the curriculum development manager for ITSM Academy. We provide it l and ITSM related education, that includes agile service management and DevOps, and lean and site reliability engineering. I'm also an author and you're gonna get a little peek into one of my publications, the ITSM Process Design Guide in this presentation.

So we're talking a lot about AI today, and I'm sure you've read articles and seen other presentations about the benefits of AI and IT service management. Very often when we talk about the benefits, we talk about how to improve a particular work activity so we can improve how we handle incidents or how we handle changes. In this presentation, I'm gonna take us in a little bit of a different direction because I'm gonna talk about how we can use AI to optimize value streams and processes.

And I'm gonna talk about it in more of a generic sense. So first of all, let's get the vocabulary out of the way. The reality is there's, you know, a massive amount of work that's occurring within your organizations as we speak.

And to try to wrap your head around all that work is impossible and really probably not at all necessary. But when you set about to improve your work, then you have to start to narrow the scope a little bit and think about how the changes that you are making are gonna affect people that are downstream from you or how it's gonna change your requirements of the people who are upstream from you. So there's this taxonomy of activities that we can work with.

It begins at the highest level with the service value chain, which at the operating model level is really how your organization handles demand from customers and the activities involved in enabling the co-creation of value with those customers. The service value chain is underpinned by value streams and value streams. Really look at the activities related to our products and services, how we design and develop and deliver products and services.

Value streams typically span multiple processes, and a process by definition is really just a set of integrated activities that supports a a particular objective. How we handle incidents, how we handle changes, procedures describe what work we do within the boundaries of a process. And procedures are underpinned by tasks, which are usually the activity, the singular activity that will be performed either by a human being or by automation in some way, shape or form.

Why does this matter? What it allows us to do is either kind of zoom out or zoom in depending upon our goals and what it is that we're trying to accomplish. So how do organizations typically go about applying AI to value streams and processes?

The most common adoption pattern is to start at the task flow. So can we automate, you know, routine activities? Can we eliminate a lot of the manual work with site reliability engineering we call the toil, right, that we have to do on a day-to-day basis.

Why do we often start there? It's the easiest, right? This is where very often there are those quick win opportunities to improve the efficiency of our work and at the same time, free up the human beings to focus on higher value activities.

So we're doing a lot of these things today, right? Intelligent ticket routing and you know, automated incident prioritization, automated root cause analysis, and in problem management, automated change risk assessments and we can apply AI to that as well. So that we do, we perform these activities in an intelligent way.

We use insights based on things like past performance, um, and, and, and past results in order to improve those activities. Process level automation then looks at things from an end-to-end perspective. So think about some of the things we're doing today.

Um, self-service provisioning, right? It's very common today where you can go onto some website and reset your password on your own. You can maybe provision a piece of software or provision an environment for a new project that you are working on, and that can happen via self-service from some kind of a service portal.

Asset discovery and lifecycle management, right? There are a lot of processes where we now have the ability to, from end to end fully automate that. And this really en enables us to have some significant operational performance, right?

We improve the speed, we improve the accuracy, we improve the consistency of those work activities. Now, sounds great, so far so good, but there is a little bit of a danger here. You may be familiar with the term local optimization.

One of the problems that comes into play when we start with task level AI optimization is that we can get really good at one task, but if we're not considering the downstream ramifications, we can actually create overburden or we can create inefficiency downstream. So maybe we use AI to automatically categorize incidents and so we're able to route incidents more quickly. Sounds great on the surface, but if AI is miscategorizing those incidents, we could be dumping a bunch of incidents on the wrong team, which results in a bunch of rework and inefficiency.

We may auto prioritize incidents based on their impact, but if we misprioritized, we could be using more resource in order to handle incidents than we necessarily need to. In other words, maybe that incident isn't as high a priority as we think it is and we just, you know, called all hands on deck to deal with an instant. So we really have to think about the end-to-end workflow.

We have to think about the requirements, the burden that we're placing on upstream activities for that task to be performed more efficiently. We have to think about what doing that task faster is gonna mean to those individuals or teams that are downstream. We have to think about resource allocation, right?

If we can speed up the flow of work to a particular team, but we don't add any more people to that team, are they going to be overwhelmed? Are they gonna be overburdened? We also have to think about things like communication and how we, if we are speeding up activities, how well we're doing, communicating what's coming and and, and what the impact is of those improvements that we're making.

Certainly we have to think about overall service quality and customer user experience. And we all know what this looks like. Any of you that have had an a situation where you needed support, you used to be able to get to, you know, a human being and they did an awesome job helping you, the organization introduced AI chatbots and now, right?

You have this technological middleman, so to speak, in order to get support. If the AI chatbot works well and you can get the answer that you need at two o'clock in the morning, awesome. If it doesn't work well and now you have to fight with that chatbot bot in order to get through to a human being, nobody is gonna be happy including that human being when they have to listen to you vent once they get, you get ahold of them.

So we really have to think about that end-to-end perspective, we have to, you know, think and, and, and work holistically to use one of our ITIL guiding principles. So here's where a value stream optimization comes into play. Value streams typically span multiple processes.

They typically span multiple functional areas within the organization. So examples might include your end-to-end product development lifecycle, right? Maybe your end-to-end customer journey, things like service request fulfillment and incident resolution, and even incident resolution.

Don't necessarily think about the incident management practice. This could be monitoring an event management detects an error, it triggers an incident. It can automatically determine is there, is this incident a known error?

Do we know how to resolve it? Do we have to raise a change in order to resolve it? And all of that activity could be perhaps performed automatically.

So let's think about it end to end. Easier said than done. You are talking now about the need to really engage with different functional areas within the organization.

You may have multiple process stakeholders that have to be involved in understanding everything that goes on within the boundaries of that value stream. Lots of collaboration has to occur, lots of breaking down of silos. Very often this means a tremendous change in the culture of the organization and the way that work gets done in the organization.

Having said that, the benefits are tremendous and you can see those benefits on the right hand side. It's often said that the white spaces between the activities of a value stream are where you reap the most benefit. So what's that white space?

This is where handoffs are occurring from one team to another. This is maybe where we've got cues where work is sitting waiting for, you know, an individual or team to start working on. Maybe this is where we've got approvals that have to occur before work can move forward.

And very often that white space, those little gaps in your, in your flow are what really significantly impact the end-to-end lead time for that particular value stream. So how do we go about doing this? There is a process for engineering processes and you see that process here very simply.

And, and, and traditionally we looked at gathering requirements. We might get requirements by doing a needs assessment by going out and doing surveys of our customers. We do some process analysis, typically this is referred to as mapping out our as is process and then doing a little bit of gap analysis.

Let's compare our as is process to the requirements of our customers and to the overall goals of our organization. Process design and implementation typically, typically involves the stakeholders for that process coming together, collaborating, trying to figure out how to close the gap between those requirements and the way we're doing things today. And then implementing a series of improvements.

Continual process improvement is monitoring performance and using metrics, for example, to continuously improve. And we begin again, right? Um, by studying the results of those improvements, we feed our requirements backlog.

So how do we apply AI to this? When it comes to requirements definition, we now have the ability to use ai, things like natural language processing, where historically we tried to have our surveys be very structured where we had a range of options so that we could create bar charts and we could very easily report on the feedback that we're getting today. We can have artificial intelligence look at what folks are saying on Twitter or what folks the um, comments, right?

The freeform comments that people are adding into surveys and be able to analyze that. It can do sentiment analysis to try to really get some insights into how people are feeling about things. Even surveys today can be as simple as, and I'm sure you interact with them on a day-to-day basis.

Give it a thumbs up, give it a thumbs down, click on one of the little range of smiley faces. And AI can analyze that information very, very easily in order to identify patterns and trends and really help to populate your requirements backlog. Now that doesn't mean that we don't still do traditional needs assessments occasionally, or we go out, we don't go out and interview our customers.

It means that we now have, you know, a new tool in our toolkit that can support us process analysis, which as I mentioned very often is looking at our as is process includes, um, a AI supported activities like process visualization and process mining where AI can actually analyze your workflows and map out the process for you. Or it can use, uh, process mining. It can, it can analyze logs in order to understand how the process is performing, where there's inefficiencies, where there's bottlenecks, and provide some insights into where there are gaps between those requirements and our as is state.

So AI can really support every aspect of requirements management and process analysis. And even if you just look at how you manage a requirements backlog, it can help to categorize those requirements, prioritize and reprioritize if your goals change. Things like risk dependency analysis, cost benefit analysis, you know, all of those things that go into building a business case, which is something we often have to do in order to justify process improvements.

AI can help with that. Learning from failure is such a big part of how we operate today. We're never going to get this right the first go around.

We don't get processes right the first time around. We're not gonna get AI right the first time around. And so really learning to formulate hypothesis in terms of the types of changes that we think we should be making, running experiments and then evaluating that those results AI can really support that type of activity.

As well as, again, process analysis, automated process mapping. There's great simulation tools out there today where you can play what if analysis that if you change the sequence of activities in a process, if you eliminate a particular handoff, what's that gonna do to your end-to-end lead time. It can, if nothing else, help you identify the bottleneck so that you are really focusing on that part of your process or value stream that's, that's causing pain as opposed to making lots of improvements that aren't necessarily gonna move the dial end to end-to-end in terms of process design and improvement, I mentioned the idea that we can use AI to do simulations of our processes.

We also can use AI to drive workflow optimization, and then there's continual process improvement. So here's the monitoring and feedback loop. We also can use AI to suggest improvements to our process models.

When we talk about continual process improvement, it's important to talk about concepts like capability assessments and metrics and capability assessments. Really look at, you know, we often use the term maturity, the maturity of your processes, but how capable are you of at performing this particular process at a level that's required by your organization? And typically when you're doing capability assessments, you need to provide evidence.

AI can help us to make sure that we're capturing the evidence that's needed for those capability assessments and that we're doing a good job of capturing and, and then evaluating the data that's needed for our performance metrics. So AI can support every aspect of process design and implementation and continual process improvement. It can begin with simple process automation.

Again, automating those repetitive or manual tasks, process or orchestration looks at shifting the sequence of those tasks. And then process optimization can look at the data and use those, the, that performance data in order to suggest improvements. One of the other real benefits of AI is one of those things that folks in it tend to hate, which is the documentation aspect of things.

So it really can also help with documentation updates. And so it requires you to kind of think about where your process repository is gonna be. Historically, we might have produced a process definition document, but today, very often we're using things like SharePoint or Confluence where we have then that capability to update that documentation in a more dynamic way.

And then continual process improvement is where we learn to improve by using data driven insights. And we do that as much as possible real time. So now we can feed the requirements backlog and we can start the cycle over again.

So how do we do this? How do we turn AI potential? Because it all sounds great, right?

But how do we turn it into reality? First of all, we have to accept the underlying assumptions and I threw out a lot of these assumptions as I went along. It's highly dependent on high quality data.

Are we capturing sufficient, accurate, clean data? We need to find performance indicators. We have to understand what it is we're trying to improve here.

Are we trying to improve lead time or cycle time or resolution rate, right? What are those indicators that we're going to be looking at to know if our improvements are successful or not? We have to understand what a process is and what a value stream is, and we have to have automation ready processes, which means they're repeatable.

If you have an organization where everybody's kind of doing their own thing, you have lots of ad hoc activities going on, um, from one day to the next, individuals might be doing their work in different ways, you're not going to see as much benefit initially, right? We want to move to a more standardized process approach. Now that doesn't mean there aren't gonna be some aspects of what we do that involve experimentation and learning, right?

That involve innovation and trying to come up with new ways of doing things. But we want a percentage of our work to be standardized because then we can let technology take care of that work. We need strong data integration across systems.

We need continuous, real-time data flow. So if you still have a lot of manual activities and you're still relying on human beings to enter in tickets at the end of the day, again, you're gonna, you're, could you reap benefits? Yes, you won't reap as much benefit.

And we need effective knowledge management. And that has not only to do with how we manage our documentation and you know, our intellectual assets within our organization, but it also has to do with how we share experiences in terms of how AI is performing. So that's sort of the technological side of things, but we also have to understand the human aspect of things.

We have to be open to change. Every aspect of work is disrupted. When you start to introduce ai, you're changing the way people do their work.

You're changing perhaps where and when work gets done, you may be eliminating some work altogether. You need supportive human oversight. So you need to make sure that the decisions that are, that are, that the suggestions that are being made AI are, are, are given critical thought.

We're ensuring that the work that's being the AI driven activities are being done in an ethical way, that they're aligned with our organizational goals. And we also really need that human-centric approach, making sure that at the end of the day, this is gonna benefit our customer and human experience. We need strong cross-functional collaboration.

Because if you change work in one area of an organization, it's invariably somewhere downstream gonna affect another part of the organization thinking more holistically. And we have to learn to trust AI insights and recommendations. But having said that, there has to be transparency in AI decision making today.

Leadership sometimes prides themselves in making dis gut-based decisions as opposed to data driven decisions. Or perhaps they have some experience that will result in them second guessing a recommendation by ai. And that's okay.

We have to apply critical thought. But in the same respect, just to go back to the top option, we also have to be open to change. And we have to be, there has to be a willingness there to embrace new ways of working.

So focus on continual improvement. Make sure that you are clear on your goals for your organization. Think about your organization's circumstances and needs.

Why are you thinking about introducing AI into your, um, organization and, and specifically in the context of improving value streams and processes. How's this gonna support your goals? Understand the capabilities of your existing tools.

And I always like to say that because I think there's this tendency among IT people to always think we need a new tool. But the reality is in the IT service management world in particular, a lot of the tools today have great AI capability, even tools that are really geared to smaller or mid-sized organizations. So companies like, you know, fresh service and caid manage engine, you know, they have great I AI capabilities.

So you don't have to have ServiceNow or BMC in order to get this type of capability. The question is, are you utilizing the capability of your existing tools? You can launch data quality and knowledge management initiatives and you know, that's something that you can start doing today and it's going to support your ability to be successful as you move forward with your AI initiatives.

Invest in upskilling and re-skilling programs. Invest in automation and data integration. Uh, activities.

Stop with the tool wars. You know, there's this age old debate of like one part of the organization's using Jira and another part's using ServiceNow. You know what?

Those tool tools can talk to each other. There's these things called APIs, right? So invest in data integration rather than kind of forcing people to maybe only occasionally use a tool that's really cumbersome for them, or making them move altogether off of a tool that they're quite comfortable and productive with onto a new tool.

Unless your aim is to move to a a, a platform approach, think about integration and then encourage experimentation and learning. And I, I said this before, learning from from failure is such a crucial part of the journey that we are on from ai. If we formulate an experiment and it doesn't work, what went wrong?

Was it a a hypothesis? Was it the data quality? Was it something about the automation?

Was there some human aspect of it that we didn't take into consideration? When we think about the importance of being lifelong learners and the role that learning from failure plays in that process, it really helps individuals and organizations to be able to continually improve. We have to consider the human experience.

We have what we see on the left hand side today, right? AI is causing a lot of fear, a lot of anxiety, right? Frustration, lack of control.

Sometimes we don't know what's going on. There's a lot of skepticism out there. What we want is on the right.

We want people to be excited, we want people to have a sense of curiosity. Notice that skepticism appears on both sides of this spectrum, and that's okay, right? You really do it.

It is very important that you apply critical thinking and that you be willing to challenge what's happening with ai, the insights that it's providing to you. The suggestions, do they make sense? Are they relevant to what you are trying to accomplish?

That's the human aspect that we don't ever wanna lose sight of. But we want people to be excited, we want them to feel confident. And I think here's where experimentation and learning is.

So, so much an important part of this conversation. Learning to say, I'm going to conduct an experiment. Here's what I think I'm going to learn.

If that experiment, if that hypothesis doesn't play out, what did I learn from that? And have the confidence to then go into the next round of improvement with that additional insight and, and, and that hopefully confidence that's needed to progress forward with the next round of improvement. Be empowered, right?

Look at what can you do on a day-to-day basis to leverage AI in performing your tasks. How can you influence your team or other stakeholders in the processes that you're involved in to, um, improve more of the end-to-end perspective. So start now.

AI can support every phase of process of the process. Re-engineering lifecycle. Where you start, one of our idle guiding principles is start where you are.

Right? Where you start is really gonna depend on the goals and circumstances and needs of your poor particular organization, right? What are your business goals and what are the pain points?

And think about this from a strategic perspective, right? What are you trying, what is the greater organization trying to achieve to achieve? You have to think about things like organizational maturity and readiness.

Again, if you have a lot of ad hoc processes, you've got a lot of manual processes, you may need to start there in just in terms of establishing a process oriented culture in your organization, a culture that values the capture of the data that's needed in order to evaluate the performance of those processes. And you also have to look at resource availability. And here's where, let's circle back to talking about, you know, AI driven task optimization.

You know, you may not have the budget, the human resources, the technological resources in order to do some of what you wanna do. But every time you automate a task, you free up human resource. Every time you learn how to better utilize your existing tools, you gain more technological resource.

Once you start to learn to talk about and articulate the benefits of what you're doing, you may find that you free up some budget resources, right? You're able to now justify more improvement initiatives by providing data driven insights in terms of what the benefits of your improvements are. So let's accept AI for what it is.

It's a tool like any other tool. It is very quickly becoming ubiquitous. It's, you know, it's becoming no different than the email and the Slack channels and the team channels that we're using on a day-to-day basis.

For many of us. It's something that we use throughout the day in many, many different ways. But like any tool, its effectiveness is gonna depend on the skills and the intentions of the humans who deploy and use it.

So good luck with your value stream and process optimization efforts. Enjoy the rest of the day.