Why Process Optimization Alone is No Longer Enough in the Al Era

When we talk about process re-design, the first thing that comes to mind is traditional process mapping techniques, like value stream mapping, process mining or maybe event storming. These are all great techniques to map out the current state of the process, but what about the future state?

This article explores how to approach process redesign to reimagine and implement your processes in an AI-native way.

Mapping the current state

To assess the current state of any process, my go-to tool is Value Stream Mapping (VSM). VSM mainly helps make the entire process visible, including steps, timing, actors, painpoints, waste and handovers. VSM gives me exactly the foundation I need to clearly identify and understand where, the current process can be improved, but not directly where AI can have an impact. Standard VSM mostly remains focused on optimizing the existing process. For an AI-native redesign, you need to take it one step further.

A good practice when mapping your current process is to look not only at the standard elements but also at where AI could be applied and how data flows through the process. Handle the following questions as suggestions to add to your standard VSM to further align it with the application potential of AI.

In general, when mapping the current state, you want to combine current pain points and inefficiencies with the potential of the new technology.

Imagining the future state

Traditional VSM maps the future state of the process more from the current state, altered for changes made to improve specific problems and pain points in the current process.

When you redesign, or rather, reimagine processes to be AI-native and address current problems and pain points, you are actually designing a whole new flow. This is reinforced because AI-native thinking brings inherent new possibilities that were not possible beforehand.

More generally, process reimagining also implies a more fundamental shift in how we look at and think about processes. This is because the process will become more of a collection of connected, potentially reusable AI capabilities than of activities. If you look at it this way, it’s easier to think AI native, and it also changes the way you think about how to judge if the process is actually performing. Just think about it, does the touch time of an activity actually matter when it’s performed by an agent? A more telling signal might be token cost per output. 

How to transform from current- to future state

The big question then becomes how to move from the current state to this re-imagined future state. Do you redesign the process all in one go, or do you take a more step-by-step, iterative approach? Both have their place, but they also come with very different risks.

A big-bang redesign can be powerful when the current process is clearly broken and hardly functional or performing, when there is strong executive sponsorship and when the organization is ready to accept a more radical change. It creates momentum and it forces people to look beyond the limitations of the existing way of working. The downside is that it can also become too conceptual too quickly. People may understand the ambition, but not yet see what the impact will be on their work. Another issue might be that the entire process would have to be trained and tweaked to be at least as performing as the current process. This can be a difficult process to get right with a larger scope of the entire process and will demand attention of specialist over a longer period to get it right

An iterative approach is often more practical and achievable. It allows you to start with a concrete part of the process, learn from it, test the new AI capabilities in real use, and gradually build trust. This is especially important because AI-native processes are not just process changes. They also change decision rights, roles, responsibilities, quality controls and sometimes even the customer experience itself.

From a change management perspective, this means you should not only design the new flow, but also design the transition. People need to understand why the future state is better, what will change for them, what will stay human, and where their expertise is still essential. If that story is missing, AI quickly feels like something that is plays into the age old change saying: people want to change, but don’t like to be changed.

From a customer perspective, the question is slightly different. The future state should not only be more efficient internally. It should also create a better, faster, more consistent or more personal experience for the customer. If the redesign only removes internal effort but makes the experience colder, less transparent or harder to understand, you have probably optimized the wrong thing or the wrong direction.

And lastly there is the human-in-the-loop perspective. In many AI-native processes, the human role does not disappear. It shifts. Humans move from doing every step themselves to setting direction, checking exceptions, judging quality, handling ambiguity and taking responsibility for decisions that matter. This is where risk management becomes part of the design, not a control afterwards. You need to be explicit about where AI can act autonomously, where it should recommend, and where a human decision is mandatory.

Hints and tips from the field

One thing I see in practice is that people often confuse automation with AI. Automation is usually about making an existing step faster or cheaper. AI can do that too, but its real value is often in changing the nature of the step itself. Instead of collecting information, interpreting it and then writing a recommendation, an AI-native flow might generate the first insight immediately and let the human focus on judgment and context.

That is why I would still recommend starting from the current pain points. Not because the future state should be limited by them, but because pain points give you a useful entry point. They show where energy is leaking out of the system today. They also help people connect the AI-native redesign to something they already recognize and care about.

Another helpful practice is to give every actor in the process a clear perspective on their future role. What does this mean for the employee, the manager, the expert, the compliance officer or a developer? AI-native redesign can easily become too abstract if it only talks about agents, models and capabilities. It becomes much more real when people can see how their role changes in the new flow.

Finally, expect to tweak. A lot. AI-native processes are not designed once and then implemented perfectly. You will need to test prompts, data quality, decision rules, handovers, guardrails and exception handling. The goal is not to make everything perfect before you start. The goal is to make it good enough and safe enough to try, learn and improve.

First steps forward

A practical first step is to select one process where AI-native redesign can create visible value. I would not always start with the most exposed or most critical process. Start with something meaningful enough to matter, but safe enough to learn. You want a process with real pain, available, decent-quality data, and enough room to experiment.

Then create a change story. Do not start just because the technology is there, or because everyone feels they have to jump on the AI tech wave. Explain why this specific process improvement initiative matters, what the future state could make possible for the company and the people involved, and what people can expect during the journey. A good change story makes the redesign less about tools and more about ambition, people and perspective.

And find the right mix of capabilities. This work asks for process analysis, AI-native redesign, iterative development, change management and deep technological skill. That combination matters. If you only bring technology expertise, you risk building something that does not fit the organization. If you only bring process expertise, you risk optimizing yesterday’s process 10% better. The value is in putting both on the table.

This is also where external help can be useful. Not because an outside expert knows your process better than you do, we always co-create on this just because of that, but because they can bring the patterns, questions and technical understanding needed to challenge the current logic. At Xebia, this is exactly the kind of work where we combine transformation experience with AI engineering and change capability.

Conclusion

Re-imagining your business processes in an AI-native way is so much more than doing VSM and eliminating waste. It is a multidisciplinary exercise and a paradigm shift in how we think about processes, roles and performance. The future state will not simply be a cleaner version of today’s process. It will often be a different way of creating value. So start reimagining today, not just to improve the current state, but to create the future state for tomorrow.