2024: Will Generative AI Reboot Your Business Processes?

What a year! Things happened at such a pace that it's hard to remember how things looked like last year.

With so much evolving and so quickly, few Business Process Management (BPM) pundits noticed the 30th anniversary of the publication of Reengineering the Corporation, a book that laid down some of the keystones of business process redesign.

Talking about round-number anniversaries, this year was the 20th anniversary of a Harvard Business Review article controversially titled "IT Doesn't Matter" by Nicholas Carr .

Twenty years on, this headline hardly aged well. It's reductionist to summarize an article in a sentence, but if you allow me for a second, Carr somehow argued that IT had matured (in 2003!) to the extent it no longer gave companies competitive advantage. Some thoughts in that article still stand, arguably not the headline.

A few months later, Howard Smith and Peter Fingar retorted with a book titled "IT Doesn't Matter - Business Processes Do". That title aged a bit better. Business processes do continue to matter, and will continue to matter for a long time. But IT still matters, too, and will continue to matter for donkey's years.

As a business leader, if you don't have a strong, innovation-driven IT function, you might as well skip the next couple of decades.

Like other fields, the BPM discipline is poised to be heavily transformed by the rise to maturity of GenAI. The question is how and when?

To be clear, the BPM discipline covers many capabilities. In "The Six Core Elements of Business Process Management", Michael Rosemann and Jan vom Brocke ponder that a successful enterprise-wide BPM practice must cover six capability areas: strategic alignment, governance, methods, tools, people, and culture. Below. I focus on the "methods" and "tools" areas, without intending to diminish the other capability areas. For more on the other capabilities, ask the "New Process Guy", Mirko Kloppenburg , who recently published a summary of the state of affairs in BPM, with a focus on strategy, people, and culture.

From the perspective of methods and tools, the main development in the field of BPM in the past couple of decades has undoubtedly been process mining.

Process mining has made it possible for business stakeholders to discover their processes from data stored in their enterprise systems (CRM, ERP, ticketing systems, etc.). It enables us to trace down which steps, which handoffs, which sources of waste, are having an impact on our KPIs. Not less importantly, process mining makes it possible to discover high-fidelity simulation models from data. These simulation models serve as "digital twins" of the process. They allow us to answer what-if questions such as: By how much would we reduce our complaint-to-resolution times if we automated the verification steps, or if we allocated more resources to the investigation steps?

The field of data-driven simulation has evolved tremendously in the past couple of years. By now, simulation models discovered out-of-the-box using advanced process mining platforms are good enough to pinpoint multi-million-dollars improvement opportunities in large and complex operations.

Process mining, in combination with machine learning methods, also makes it possible to deploy predictive and prescriptive monitoring dashboards, for example to detect and preempt negative case outcomes, such as customer complaints or SLA violations, way before they occur – more on this in my previous post on these topics.

But for all the support it gives us during the discovery, analysis, and monitoring phases, process mining is limited when it comes to a key phase of the BPM lifecycle: the process redesign phase. It's in this phase that, having identified and analyzed a set of weaknesses, issues, or opportunities in a process, the business stakeholders come up with concrete improvement options, and assess these options to determine how to bring up the performance of a process with respect to relevant KPIs.

While process mining assists users in evaluating options during business process redesign, it still requires the business stakeholders to figure out what changes to make to address an issue or to exploit an opportunity. Process mining tells us, for example, that we have high rework rates for certain types of cases, or that workers spend considerable time handling cases that are later cancelled (overprocessing). Where process mining falls short is when addressing questions like:

• Which incentives or which work instructions do we need to change to slash down those rework rates that are causing us KPI headaches?
• Which changes in decision logic or policies are needed to reduce compliance violations while still giving enough autonomy to the on-the-ground workers?
• Which customer touchpoints should we consolidate to reduce overprocessing?
• How feasible is it, given our current org structure and policies, to do this change, or to do that change?

For good or for bad, these questions require us to heavily rely on tacit domain knowledge, creativity, beliefs, and stakeholder buy-in. There are many toolkits and methods to guide us through the business process redesign maze, see for example those in Lean and Six Sigma, or the various redesign heuristics used in the field of BPM. (Again, talking about round-number anniversaries, it's been ten years since I had the pleasure of co-authoring the book Fundamentals of BPM with my colleagues Marcello La Rosa , Jan Mendling , and Hajo Reijers , which includes a chapter on business process redesign.)

So, until now, the business process redesign phase of the BPM lifecycle is the odd one that is hardly supported by IT tools. It's the phase where domain expertise, intuition, and guesstimates dominates.

Meantime, after slowly cooking in the background, 2023 has been the year when the lid went off on generative AI. Many pundits are attributing magic powers to GenAI tools, including controversial arguments that GenAI can have "reasoning abilities".

Magic powers aside, the burning question for BPM practitioners is: Can GenAI help us transform or optimize our business processes in the year to come?

Diving into GenAI use cases in BPM, I ended up classifying them into two buckets, following Henry Ford's famous quote:

If I had asked people what they wanted, they would have said faster horses.

Some use cases are about using GenAI to make faster horses. These include using GenAI to generate Extract-Transform-Load (ETL) scripts to accelerate data preparation for process mining or other analytics use cases. Others are using GenAI to generate queries to answer questions about a process, such as finding the effort spent in rework loops for different types of customers. As it turns out, these are use-cases that business stakeholders are already tackling using process mining tools. Deploying GenAI in these use cases can help to reduce the effort in tools where heavy data engineering is required to build a process mining analysis. It's fundamentally about "faster horses" (and sometimes the accelerating effect is marginal).

Other use cases have higher value potential, because they allow business stakeholder to tap into new information sources during the discovery, analysis, or monitoring of business processes. These are the "better-than-a-horse" use cases. One such use case is semantic preprocessing of textual process documentation. For example, GenAI makes it possible to determine which elements of a textual work instruction or in a policy document, relate to a given activity in a process. This allows business stakeholders to dig deeper into steps in the work instructions or guidelines in a policy that are having a direct effect on KPI violations. Now, we're talking about making something possible, that was not within reach previously, because it required so much manual tuning that you would not do it for every step in your process and/or for every document in your knowledge management system.

But where I see the greatest impact from GenAI is in the business process redesign phase. I see at least two use-cases in here.

First, GenAI makes it possible for business teams to generate improvement options, by combining relevant documentation (work instructions, policies, process models, etc.) with insights extracted from a process mining platform, to generate business process improvement options. For exampl,e, given a time-to-resolution KPI, currently violated in 15% of cases, a business analyst can ask "How can we slash KPI violation rates by 10 percentage points at constant resource cost?". I call this use-case the redesign discovery use case.

Second, GenAI in conjunction with simulation, enables business stakeholders to explore what-if scenarios conversationally, by asking questions like: What reduction in SLA violations would we get if we achieve an automation rate of 20% or 25% on one or more activities? What would be the impact on wasted effort? On waiting times? This is about using GenAI to assess improvement avenues, in other words, this is a redesign assessment use case.

Third, and not least, GenAI allows business stakeholders to identify which additional questions or factors they may need to consider when making changes to a process, for example, which compliance rules they need to consider. In other words, we see GenAI helping business stakeholders to determine which questions they should be considering, rather than just answering the questions posed by the users. In essence, this is a redesign enhancement use case.

Now, let me come back to the question: "Will GenAI Reboot the Business Processes in your organization"?

Eventually, the answer is a YES.

However, how fast it happens will depend on which use cases we prioritize and emphasize when deploying GenAI in our BPM practice. As a BPM practitioner, spending some time in the "faster horses" use-cases can help you get some grasp on the technology. In the medium to long-term, though, what will make a difference are the "better-than-a-horse" use cases, and among them, the business process redesign use cases: redesign discovery, redesign assessment, and redesign enhancement.

Some might ask if there are also more compelling application of GenAI in the business process implementation and execution phase? Surely yes. Besides the many opportunities that GenAI provides to process workers to enhance their efficiency, particularly when handling customer interactions, there is a lot of potential to use GenAI to trigger runtime interventions during the execution of a case, for example when a case is predicted to lead to a customer complaint or an SLA violation.

It's clear GenAI will unleash a lot of horses in the field of business process improvement. Some of them will accelerate the discovery, analysis, and monitoring of our business processes. Others will enhance our ability to drive transformation.

Further Readings

• Katsiaryna Lashkevich in Can ChatGPT heat up your business process redesign, shows how LLMs can be used to generate business process redesign ideas, ranging from "incremetal redesign" ideas to "transformational" ideas.
• Ihar Suvorau in GPT-4: A Paradigm Shift in Business Process Improvement, shows how improvement ideas generated by GPT-4 can be validated via simulation models.
• Urszula Jessen , Michal Sroka and Dirk Fahland in Chit-Chat or Deep Talk: Prompt Engineering for Process Mining, show how to tackle a range of process mining questions using LLMs.
• Alessandro Berti and Mahnaz Qafari in Leveraging LLMs for Process Mining, also show how to design prompts and data abstractions for answering process mining questions with LLMs.