On the way to their digital future, insurance companies are considering different solutions and tools for business process management. The main promise of this kind of transformation includes digital workspaces and paperless offices. However, more ambitious Insurers involved in digitalisation initiatives should aim much higher. They should analyse and try to automate their business decisions.
Of course, fully managing a business process with a computer is not without benefits of its own. One can anticipate at least lower costs, less space requirements and reduced impact on the environment. However, transforming the way we do our paper-shuffling is still marginal from a business outcome perspective – even if we put everything in the cloud. In addition, superficial digital transformation projects seldom give more insight into how the work is done and where to find further opportunities for automation.
Complementary to the business process management initiative is decision management. Organisations practising decision management consider decisions as their most valuable assets and try to make them more precise and responsive. They make use of decision management tools to model business decisions and if possible, they automate them.
In the insurance industry, thousands of decisions are made every single day within every single Insurer. Even within a simple sales process, one must decide what the optimal insurance product is for a specific customer's needs. And this is only the beginning of a chain of decisions that happen later on. Next comes risk identification, underwriting, possibly also the customisation of coverage; and all this before the policy is even written and concluded.
During the analysis phase of IT projects, business decision modelling usually starts with the identification of the key decisions the business has to make to generate the desired business outcomes. Linking identified decisions to measurable business goals makes them accessible for periodic revision and continuous improvement.
The most important part of decision identification is to articulate what it is that needs to be decided and what information (input) is needed to make these decisions. The first part is figuring out the question while the second part is about what we need to answer the question. The magic of creating an answer from input is called decision logic. As most key decisions are complex, one has to break them down into smaller decisions. Usually supporting decisions provide input for more important ones, making a web of decisions that represent specific know-how.
By using this network of interdependent decisions and required information, an organisation can identify which decisions can be automated. Identifying decisions during the analysis phase of the project makes implementation and configuration of business rules more straightforward.
An important part of business rule implementation is the process of publishing and deployment to the production environment. Changing a rule can have major effects on business outcomes. The process must be transparent, with traceable changes and, when possible, the system should allow users to test and simulate the validity and effects of the proposed changes.
At Adacta, we try to recognise positive insurance practices and patterns and embed them as recommended rules across the various business modules of AdInsure (such as underwriting rules during the sales process, all the way through to processing during the Claims process). In that way, our new clients benefit from all the lessons we have learned with our out-of-the-box solution.
When decision inputs and goals are explicitly defined, it is sometimes possible to create a business rule by fitting a statistical model with a training data set. The results of machine learning (predicting outputs based on some inputs) provide exactly what is needed to implement the decision logic.
Underwriters have been asking questions and making decisions for hundreds of years now. Modern-day machines are able to utilise the history of those decisions in the blink of an eye and streamline the process in such a way to achieve the optimal outcome, for both the insurer and the customer.
Ideally, the system should be able to convert the results of the machine-learning process directly to a business rule. The new version (and generation) of our insurance platform AdInsure is built to support the business rule concept. Most of the insurance business processes that come as part of the platform can be automated by simple business rule logic, in addition we also run several proof of concepts for more complex automation. For example, we have successfully tested the possibility of deploying a fitted model from the Jupyter environment, using scikit-learn, a popular machine-learning library.
Starting IT projects with a decision management approach may require more analysis effort than just putting organisational processes on a digital platform. However, the approach promises simplified and transparent processes with better responsiveness in operations.
Also, it gives you the possibility of data driven, self-learning business rules, which, in the long run, will bring value through both cost optimisation and business agility.