Applying Artificial Intelligence in existing Business Processes

Posted on: April 4, 2019 by Peter Kalmijn

AI means to CxO's: driving customer journeys into better and deeper experiences and highly personalized digitized processes. Business process management (BPM) is on the verge of a serious makeover. Where conventional BPM primarily focuses on process efficiency and cost reduction, the new wave is pushing for business agility, customer (business) experience and digital automation. This wave tries to use artificial intelligence in business processes where the most benefit is to be expected. The crucial question is: which business processes are the most suitable to apply AI?

Leveraging AI and machine learning surely will help businesses become more proactive, make better decisions for the future and improving the customer experience in the short term. AI in general, machine learning, deep learning and natural language processing can have a significant and positive impact on how business processes are run. But how will machine learning cope with ever changing policies and legislation? And with transparency and compliance? Can machine learning on its own deliver the business value envisioned?

The answer is clearly: no!

AI essentially thinks in a similar way to humans. But there are two distinct ways to combine to make computers do the intellectual tasks traditionally reserved for humans: 1. using Data Knowledge and 2. using Expert Knowledge.

1. Data Knowledge: Data Science

The essence of machine learning and deep learning is the learning form data and making predictions based on historical data. It delivers what we call Data Knowledge. Predictions are very valuable, because it brings us the knowledge of customer behavior and of interpretation of voice and vision. But a prediction is not a decision, and is non-deterministic, not explainable and not really audit-able. This poses problems for many companies.

2. Expert Knowledge: Business processes

This brings us to take a better look at the declarative, deterministic part of AI. This part of AI copes with the controlled part of business logic, which comes from experts knowledge, and based on experts interpretation of regulations and policies to be applied in business processes. Which is a good thing, because transparency and explain-ability is of great value to companies.

Digital Business Analysis: applying ModelDriven-AI

In my work as Digital Business Analyst, I apply what it is called 'ModelDriven-AI'. By modeling a particular decision for use inside an existing business process that is currently a manual decision by an Subject Matter Expert. Thus I elicitate the Expert Knowledge and structure into a Decision Model. Now I can clearly point out which parts of the decision are better approached by Expert Knowledge, and which parts can be done using Machine Learning, making use of Data Knowledge. Thus combining the best of two worlds, both using different AI-algorithms: Inference and Machine Learning algorithms.

Let me give an example: First, I establish a BPMN model of the Business Process which facilitates the targeted decision. This transforms the result of the decision into some action.

Next I model the decision itself, using DMN, making a Decision Requirements Model. This gives me the means to indicate for which part of the decision I choose Data Knowledge and Machine Learning, and for which part I choose Expert Knowledge and rule-based inference.

The models I derived from Machine Learning I model as Business Knowledge Model. This is the symbol to depict a function encapsulating Business Knowledge. Personally I prefer to model Data Knowledge to the right side of the model. To the left side of the model a subject matter expert is modeled to indicate the Expert Knowledge provided to the decision model. The sub-decisions however are all rule-based by default. This implies that the criteria learned in the Machine Learning models are translated into executable business rules.

How do I determine which parts are kandidates for the Machine Learning approach? During my elicitation of Expert Knowledge, I sometimes hear something like 'I don't know - I do it on guts feeling'. Exploring this a little deeper, this may very well point to Data Knowledge. The expert has seen something often, which is to reliably be found in the Data. And this is what Knowledge Machine Learning can derive.

Enterprise Decision Management

But how to manage automated and not-yet automated Decisions over time? This calls for strategic management of business decisions: Enterprise Decision Management. Apart form ModelDriven-AI this encompasses Process and Decision Life-cycle Management and Decision Architecture too.

Enterprises need to make sure to take the steps to adopt the technologies in the right order to ensure the lasting outcomes they’re looking for. To start with establishing the right Modeling Practices - focused on business decisions. Next start automating business decisions based on Expert Knowledge - thus laying the foundation for Decision Automation. And finally incorporating Data Knowledge to this decisions for very specific situations.

Answering the question "which business processes are the most suitable to apply AI?": I think all those business processes which rely on effective and transparent, explainable decision making. This includes almost all processes in all markets: finance, government, industries, health among others.

From September 10th to 12th, over 400 people including Atos technical experts, partners and start-ups will gather during our 2nd Expert convention in Madrid to build the future of technology and to instil and drive technological changes that will help our clients master these challenges in a sustainable and ethical way.

Share this blog article

About Peter Kalmijn
Business Engineer of the Digital Transformation and member of the Scientific Community
IT Consultant at Atos, Netherlands. Peter is a visual-spatial thinker with an international multi-cultural background. He has a special interest in IoT, Automated Decisioning, Enterprise Decision Management and Business Rules. And combines his interest in creative ways with over 30 years of IT experience gained with business- and software engineering. Peter authored various papers and articles and speaks at events. Additionally, he is lead-trainer of the Atos EDM related courses. He dedicates his time helping organizations with Digital Transformation. Peter is thought-leader of the Atos competence Business Information Analysis (BIA) and Guild Master of Atos "Enterprise Decision Management". He is a member of the Scientific Community and the Atos expert network with a focus on Model Driven Development, Business Information Analysis, Process Modelling, Decision Modeling and Automated Decisioning.

Follow or contact Peter K.J.