Written by GoodData Author |
In a world flooding with data, discovering insights and acting upon them can give businesses a much-needed “unfair” competitive advantage that drives higher revenues, boosts customer loyalty, lowers operational costs, and improves management of regulatory and compliance risks.
However, given the volumes and variety of data streams that are available to enterprises, it is no longer sustainable for businesses to wait for analysts to mine their data for insights. There are simply too many variances and patterns to be humanly fathomed or objectively analyzed at scale.
As an example, consider the attractive premiums that motor insurance companies offer to incentivize drivers for exemplary driving habits. Traditionally, these have been derived from customer demographics, claim history, public DMV databases and good old actuarial tables. Innovative insurers are going beyond this to determine driving habits, often represented as a score and imputed from car-telemetry data, car-maintenance data syndicated from car service centers, and traffic and weather patterns. That’s a variety of data coming from disparate sources, both from within and outside the enterprise, static and streaming that is impossible to analyze with human intervention.
And then there is the weak link between insights found and action taken. The whole point of discovering insights from data is gaining an enhanced ability to take an appropriate action. While online marketing platforms have embraced this successfully, business processes automation platforms do not bridge insights to actions, or do so as badly engineered afterthought. Investments in automation are more often spent on discovering insights than responding to them with actions that drive business outcomes.
Businesses are addressing these challenges by deploying, or are considering the deployment of, robotic automation. Robotic automation uses artificial intelligence (AI) and machine learning (ML) to automatically predict insights and patterns, driving appropriate actions with minimal human intervention. Predictive ML models created by data scientists are operationalized within the context of a business process. Robotic automation has been successfully deployed in a diverse set of industries and functions including, mortgage processing, T&E handling, purchasing, and facilities management.
But there is one issue that robotic automation solutions often fail to consider.
The Makings of a Skynet!
To better understand the scenario, we need to step into the shoes of the “business process owners” responsible for the business process that has been enabled by robotic automation. These business process owners are accountable for the huge investments made into the automation, and are expected to realize the ROI. Effectively, these executives are accountable for the decisions taken by the embedded machine learning algorithm and the automated actions taken by it. It is a frightening prospect for a responsible business process owner to be held accountable for a business process that is orchestrated by what is essentially a mathematical algorithm.
Would business supervisors allow a machine to control the future of their performance, and eventually, the fate of their careers? That’s a scary prospect, isn’t it?
Supervisory Control with System of Insights
Here’s where we need to think in terms of the solution as a System of Insight. A System of Insight approach emphasizes that the technology of robotic automation needs to be augmented with solutions for the people who deliver and maintain it.
These solutions provide answers to questions that concern business processes: How much of predictions has the predictive models actually contributed to? How have these recommendations varied with time? How has the quality of predictions varied with upgrades to machine learning models? Has it improved or become worse? And if the predictions are validated, how accurate have the predictions really been? How do measures of accuracy vary with time? And finally what is impact does the automation have on performance business metrics?
Furthermore these solutions provide the supervisors with dials to tweak the behavior of the automation, and dashboards to see trends that can be used to have a meaningful dialog with the data scientists who make the changes.
The paradigm of System of Insight ensures that Artificial Intelligence (AI) and Machine Learning (ML) are subservient to executive oversight and corporate governance, just as manually-driven processes are. Operational leadership can now safely commercialize the flood of data to drive positive business outcomes. And when conceived as a System of Insight they are assured that they are never going to run by a Skynet!
Muralidhar Thyagarajan is a Group Product Manager at GoodData, which provides an Enterprise Insights Platform that enables enterprises to create and deploy Systems of Insight. With more than 20 years of business process and workflow automation experience, Muralidhar has successfully productized robotic automation for mortgage processes that have been adopted by top-tier mortgage lenders.
Written by GoodData Author |