Companies are scattered when queried about their definition of artificial intelligence and their stance on where to take it. We first got a glimpse of neural networks in the late 80s. Then we found ourselves questioning machine’s ability to do work. Hollywood’s utopian/dystopian visions created a number of opinions for us; but in the end, we failed from a lack of processing horsepower to really embrace the AI that were in the imaginations of people and businesses. While we had some of the math, we were doing it long-hand and it took forever to get anything accomplished.
With the advent of technologies from NVidia and the work of Deep Learning, new questions became possible and we ultimately look at some of the options that the movies have shown us. By itself, the new term Machine Learning is useful to a degree, but the real need is the ability to drive decisions that create behaviors we seek. Some of these goal behaviors are the original ones: upsell; cross-sell; and driving customer brand loyalty. We believe there are other decisions which exist that can leverage these new capabilities bringing them to the front line of business processes that drive the organization regardless of mission:
- Services and Customer Support: Diagnose as an expert within the same conversation with a customer and not needing to “hand-off”;
- Insurance: Underwrite risk in seconds with a response specific for customers OR agents and provide tailored next actions to conclude the underwriting experience;
- Banking: Get credit approval authorizations in the process;
- Healthcare: Make immediate decisions for initial triage and recommend other activities to support preventative care and improve overall wellness;
- Healthcare: Optimize multiple channel claims processing to arrive at the best outcome that meets business objectives;
- Insurance: Provide case workers with decision support on claims to improve claims resolution;
- Research: Extend the ability to perform event-based analysis and review and provide guidance on additional experts in the research community whose expertise and capability for evaluating various topics can be applied;
- Government: Direct legal system case-work decisions through supporting judges in making their decisions reducing the time to determine outcomes.
With all the possibilities, it is important that we understand that not a single machine learning process is paramount, but it is instead an ensemble of processes that allow us to discern and make resolutions to arrive at the outcomes we seek. Through intensive work, we need to identify critical features and how they are engineered into workable predictors. Within those predictors, we must examine what machine learning algorithms we require, and precisely what needs are answered. From there, we can continue the chain of learnings until we conclude with a result to the question was asked.
4 distinct areas of learning exist:
- Classification: Where we need to determine the characteristics of an object and classify that object into 2 or more distinct sets that we know about;
- Regression: Where we need to look at a series of predictors and need to determine a quantity;
- Clustering: Where we have the computer take unmanaged data and try to find classifications for us. Within this, we identify how many distinct groupings we want and then the machine tries to slice the data into those sets;
- Dimension Reduction: Where as a set of data, we have far too much data and ultimately noise, so we need to examine the noise and reduce it.
Within these four key areas lies many different approaches and ultimately the art of machine learning is putting together the various types to construct a single artificial intelligence.
Within the continuum of BPM, if we can leverage these machine learning capabilities and further explore how we can reduce work effort or more effectively bring the desire of our customer closer into the alignment of our business objectives. We need to recognize that inevitably; BPM will see a new disruption from the inclusion of machine learning.
- Many of our companies embrace BPM now and we need to further look at machine learning as another means to improve the ROI on the initial BPM investment. As with other disruptive changes to a business-technical environment we need understanding that machine learning cannot be applied without understanding its strengths and potentials. Achieving the desired improvements, we need to explore a construct we term as adaptive process management. Specifically, we want to explore the actions of the business process and have the machines learn the methods of work and then allow for quick optimization of those methods into the BPM ecosystem. Leveraging machine learning for operational decision making contextualized within a process enabled by BPMS becomes all the more powerful given the insights that further inform our decision strategies from information gleaned from a given case as it navigates the process network.
From this lens, we believe that bringing this together in three parts is critical:
- Recognition of the problem we are trying to solve. In many cases, the businesses we interact with are going with the standard marketing answers, but we need to explore ways to bring different questions and those answers to our clients to assist in decisions that create extensive competitive advantage
- Utilization of our solutions bring about an approach that help to not only recognize the models and understand the key aspects of the problem, but we also bring our expertise in the particular business segment and detailed focus within that business
Leveraging the collaboration within a safe-harbor the construct that allows for our customer’s brightest minds to interact with our practice to create new innovative transformations that give us key insights into new ways of achieving better business outcomes.