Business process management (BPM) in the age of generative artificial intelligence (AI)

If you can dream it, can generative AI do it?

If you are following tech news or social media, you may be overwhelmed by the number of posts and articles dealing with generative artificial intelligence (AI) and its potential impact. Generative AI is promised to serve as a lever for a new level of productivity gains and is just about to land in the reality of companies and organizations. Popular generative AI services like ChatGPT by OpenAI, Google Bard, Anthrophic’s Claude 2 or image-generating AI like Midjourney stimulate the creativity of what could be even possible sooner or later in a business context. As a part of this ubiquitous wave, an increasing number of business application vendors adopt generative AI to their applications now to serve the needs of their users and to ride the marketing wave of a hype.

Leaders across the globe are exploring how generative AI can be used to increase competitive advantage. In the context of business transformation and business process management (BPM), identifying synergies between business processes – defining, optimizing and activating processes within and across corporate borders – and the ability of generative AI to generate content which is valuable to stakeholders in BPM – will be critical maximizing competitive edge.

WEBINAR: The End of BPM? Not So Fast! 

First, some generative AI basics

Artificial intelligence (AI) is a big field of research and generative AI a part of the domain. In generative AI, large language models (LLM) build the foundation of content creation having big data sets resulting from masses of crawled data (e.g., books, blog articles, web pages, communication etc.), and probabilities on their relations. Within generative AI services, they are complemented by huge amounts of training data on what are good or bad responses. In the past months, we have observed some progress with updates on specific generative AI services (e.g., from CHatGPT 3.5 Turbo to ChatGPT 4) increasing the likeliness of getting good responses, for instance with the improved ability of reasoning.

The interaction with a generative AI model is prompt-based, meaning a natural language-based request is given to a system in a spoken or written format. The response from generative AI is provided in either textual written or spoken form (e.g., aided by AI-based speech synthesis) or in the case of images in a graphical output format. The output of generative AI services with generic LLMs can be used as they are but for sure then only provide content from the generally available crawled data and training involved. Thus, it is key for many use-cases for generative AI to make it working for more special and dedicated scenarios e.g., with adaption, fine-tuning and parametrization.

Business processes and generative AI – what could be possible?

While it is highly tempting to think you could and should adopt generative AI in almost any space, it makes sense to focus and pick the use-cases that have the biggest impact on your business.

So, it is wise to understand your needs first. Generative AI is doing nothing else than executing content-generating tasks of your business processes (e.g., write or conclude an article or post, create an analysis, configure a system) which used to be done before generative AI exclusively by people.

When looking at potential areas of application, there are three critical questions to be asked:

1. What is the technology generally capable to do – now and in the foreseeable future?
2. To which areas of BPM and/or business transformation does the generative AI technology apply best?
3. How can generative AI add value to BPM/business transformation software and vice versa to achieve synergies?

What the technology is capable of doing

A recent article from McKinsey predicts expected value creation in the areas of marketing and sales, operations, IT/engineering, risk and legal, HR and utility/employee optimization. Key purposes are drafting, summarizing, complementing, optimizing content (e.g., articles, code, documents, images) and supporting chat-like conversations between people – for instance employees or customers – and machines with a knowledge base in the background. These are popular tasks that generative AI can execute or at least prepare to a large extent. A customer of ours recently described the output of generative AI like the one expected from a junior consultant: Collected and prepared content – like summaries, presentations, calculations, etc. – based on ad-hoc researched data.

Areas of interest for use-cases around BPM

Looking at the different areas of application for BPM software, we see that generative AI can act as a helping hand for the BPM user working on a specific task.

BPM software + generative AI = match at first sight

Incorporating AI into BPM software is not a completely new phenomenon: innovations like AI-based translation for process descriptions or the AI-based Root Cause Miner for process mining performance indicators are already available to BPM users.

A key task in BPM is working on a proper and efficient way to draft a process.

This could be achieved for instance demonstrating how easy it could be to integrate content generated from one of the popular generative AI services and to work with draft input that can be adjusted accordingly by an expert. No magic with this use-case, however highly valuable – not every domain expert is a natural born writer of content that is supposed to be published to the whole organization. Many people seek perfection and then never get the task done. Having a draft generated by a machine which is good enough to get started is a welcome productivity gain for everyone.

But generative AI-driven capabilities in BPM could go easily beyond this idea: what if I or any other (non-)domain expert can ask a BPM software to draft a process with minimal input? What if I can handover written documents like standard operating procedures (SOPs) to a generative AI-enabled system and get a well-documented process map (like BPMN or EPC) out of it as a result? If you ask generative AI to create a process as a .bpmn file that can be imported into a BPM software, you get a very good result already. 

The future of generative AI and BPM is just about to get started

A lot is happening in generative AI. In the BPM space, there are a number of interesting applications, especially at those steps in which resources are scare or adoption fails. With generative AI, content can be generated fast, easy and by users without any special expert skills. Using natural language to run a task might feel a lot easier than learning and navigating a system.

The main lever is the ability to decouple the availability of knowledge or expertise from the execution of knowledge-intense tasks. With generative AI, big amounts of generic and specialized knowledge are accumulated in a way and made centrally available through a service provider that can add big value for BPM. As a result, everyone can be(come) a process expert – at least in the early stages of a project. This is for sure not restricted to a specific area of BPM like the design of processes, it applies to the full life cycle of processes: from insights to plans to actions to results.

As there is a lot of hype around this topic I will end with some advice: do not trust anyone who promises an easy solution to a hard problem. Having a great outcome always requires some great work such as developing the right feature in the right manner on generative AI in BPM or getting a great outcome (e.g., a new process definition) that makes a difference in customer-centricity, quality, cost, or time.

Similarly, to the availability of stock images that are used widely nowadays in marketing and presentations, using generative AI can be a cheap and quick lever for getting started but it will not be the holy grail for the result. But getting started is better than not getting started.