Decision Automation – The Road to Real AI Deployment

Introduction – process automation and decision automation.

Efficient processes and effective decisions are the building blocks of a successful organization. Therefore, process automation and decision automation have emerged as the key building blocks of the digital enterprise.

Process automation uses technology to automate complex (and often manual) business processes. It typically has three functions: automating processes, centralizing information, and automating data input (data capture). It is designed to remove bottlenecks, reduce errors, remove manual interventions, and prevent loss of data, all while increasing – transparency, communication across functions, and speed of processing.

Process automation is the first step in the digital journey of an organization. It also enables extensive data capture and reduction of data quality issues which forms the foundation for all Artificial Intelligence and Machine Learning applications.

Process automation enables enterprises to reap the initial benefits of digital transformation. However, to achieve the next milestone in the digital journey enterprises must automate the decision-making process which is often manual and requires the “implicit knowledge and expertise” of experts within the enterprise. Decision automation solutions provide a recommended decision. They also provide reasons because of which a specific recommendation is being made. Both the recommendation and the reasons are important for integrating a decision automation solution within the business process of an enterprise.

Decision automation helps enterprises make better, faster, error-free, and people-independent decisions thereby allowing exponential growth while ensuring compliance, minimizing risks, and reducing cost of operations. Decision automation uses artificial intelligence (AI), decision keys (data/metrics/indicators etc.), and business rules to help organizations automate the decision-making process. Decision automation is the foundation that enables adoption of AI and transformation of business processes and products. It acts as the most important lever to accelerate the digital journey.

Types of decisions and scope of decision automation.

Decisions within an enterprise can be classified under these three broad categories:

  • Long-term strategic decisions (example – if a bank should launch a new product).
  • Mid-term tactical decisions (example – if the bank should start targeting customers with limited or no credit history).
  • Frequent-operational decisions (example – if the bank should approve a particular customer for home loan).

Automation provides the maximum benefit for frequent-operational decisions; particularly ones which require the “implicit knowledge and expertise” of experts. Decision automation in such cases involve:

  • Business rules that replicate the thinking process of the expert.
  • Machine learning models that capture pattens from past data. These models may be even more accurate than the human expert. Or at least comparable to the human expert.

Even for mid-term tactical decisions, automation assumes critical importance and is often used by experts as a guiding tool. In this case, the reasons for the recommendation are often used by human experts to understand the information landscape and decide if the recommended decision should be over-written.

Figuire-1: Types of decisions and scope of decision automation.

Importance of decision automation.

Decision automation helps enterprises make better, faster, error-free, and people-independent decisions thereby allowing exponential growth while ensuring compliance, minimizing risks, and reducing cost of operations. Manual decision making is often the biggest roadblock in scaling the operations of an enterprise. Manual (or human-driven) decision making requires many experts who can apply their implicit knowledge in a consistent manner. Such consistency is often very difficult to achieve, leading to non-uniformity in business outcomes, customer dissonance and lack of compliance to established processes.

For example, if a bank or an insurance company uses manual underwriting to evaluate every new loan or insurance application, then it will require many trained underwriters who will apply the same underwriting rules consistently; soon this becomes the most important bottle neck for scaling the operations of the enterprise.

Similarly, if a bank with a large credit card portfolio, attempts to use manual review to increase credit limits of existing customers, then it will be almost impossible to manage this operation given the volume and the frequency at which such review must be performed; and the variety of information that must be evaluated to identify the new limit for each customer.

If a retailer attempts to develop a program of extending special discounts at the point of sale, based on customer value and risk of dis-engagement, then it must train thousands of store clerks which is costly and time consuming.

Decision automation is the mainstay of digital business models like FinTech, Insurtech, eCommerce etc. These business models are built around low-touch digital processes, promising super-fast turn-around time, hyper-personalization, and extremely low cost of operations. Hence, automated decisioning is the only mantra for the success of these models.

Adoption of AI is a key focus of almost every enterprise. Implementation of AI (using machine learning models as the backbone) necessities the adoption of decision automation along with a strong focus on re-imagining products and processes. To scale the mountain of AI adoption, many enterprises are looking at pre-built AI models. Such models must be encapsulated within a decision automation framework so that they can provide the quick adoption that enterprises are looking for.

For example, if an enterprise is attempting to adopt a pre-built machine learning model to allocate marketing spend across physical and digital channels, it is not enough to have such a model in place. The enterprise needs a complete decision automation solution that will provide connectivity to all systems to obtain the required information to run the model. It will also require a business rules framework that will allow marketing managers to add additional overlays on top of the pre-built machine learning model (example – enhancing the spend for billboards if a new celebrity has been recently engaged as the brand ambassador).

Return on investment from decision automation.

Measuring the return on investment (ROI) is important to determine the value addition from investments in software, model development and process-reengineering that are required to implement a successful decision automation solution. Exact ROI calculation can depend on the metrics that matter the most for an enterprise. However, the following broad metrics are relevant in most cases.

Better quality decisions.

Automated systems tend to make more consistent and bias-free decisions, and avoid human errors, fatigue, lack of knowledge/expertise etc. In many cases, machine learning models that are part of the automated decisioning process, may outperform human judgment.

Figuire-2: Drivers of ROI for decision automation.

Increase in efficiency.

Automated decision-making often increases process efficiency and reduces overall turn-around-time for business processes. Many innovations in business processes like 5-minute approval of loans and insurance, real-time recommendations, automated inventory replenishment etc. have been achieved through decision automation using machine learning models and business rules. Such benefits not only help reduce operating costs, but also help in faster scale up of new businesses and enhancement in customer satisfaction.

Better quality decisions.

The quality of a decision is measured by the impact on performance metrics like credit loss (delinquency), new customer activation, offer take-up rate, collection efficiency etc. The effectiveness of the decision depends on the depth of understanding about the underlying process that drives the outcome. Such understandings get automated effectively using a combination of AI tools and human subject matter expertise. While the models can be developed from past data, human judgement must be codified as decision rules which often are derived from experience and business knowledge. Decision automation solutions enable effective combination of the two; thereby often outperforming human decision making or pure machine learning model driven decision making.

Increase in employee productivity.

Increase in efficiency naturally helps in freeing-up employee bandwidth and thereby enhance productivity. In cases where the recommendation of the automated decisioning system is used by a human for making decisions, even in those cases, automated solutions help in reducing the time required by the human to make decisions.

Reduction in errors.

Automated decision making enables enterprises to apply a given set of rules or models in a consistent way, thereby reducing potential human errors. Decision automation also enables the enterprise to generate alerts and notifications in case some input information seems anomalous or if there is need of human review. Automation ensures that the system never fails to fire relevant alerts and notifications.

Consistent decisioning.

Automation enables the enterprise to eliminate the potential of human bias and the possibility of different units (branches, stores, warehouses etc.) adopting a different approach with regards to the same types of decisions. This ensures consistency in customer experience and adherence to prescribed standards. However, in some cases there may be a real need to customize decisions different units within the enterprise. A decision automation system ensures that enterprises consciously create business rules which include such customization, rather than leaving it to individual “human” decision makers.

Regulatory compliance.

Increasingly regulators are asking enterprises (particularly banks, lenders, and insurance companies) to have the capability of informing customers about the criteria that was used for making a specific decision with regards to credit approval, pricing or even marketing offers. Adopting a decision automation framework helps in providing such reason codes. Decision automation also helps enterprises in creating extensive logs of past decisions along with an objective criterion for evaluating a decision rule or model. These are key elements for satisfying existing and emerging regulatory requirements. 

Leave a Reply

Your email address will not be published. Required fields are marked *