Embedding intelligence in enterprise applications: Objective, efficient decision making
By : Naveen Chandra
AGM- Product Management
Enterprise software applications traditionally are designed for executing operations, however decisioning mostly need intelligent manual intervention. This approach places very high value in the experience of individuals, thereby bringing a greater deal of subjectivity into the system, which can be avoided. With the advanced utilisation of big data, machine learning and advanced business analytics close to 90% of the subjectivity involved can be reduced in operational decision making. Further, efficient systems and applications create well-oiled execution paradigm. However currently the reality is far from ideal. Large enterprises are creating Data sciences teams with the objective of bringing in greater deal of intelligence into their systems and workflows. However the disconnect between business and data science teams often leads to execution silos creating non permeable situations for data based insights to flow across the Enterprise.
Enterprise applications typically capture the transactions but are not designed to capture the knowledge and experience which can take them to next level in decision making. Packaging right kind of intelligence within Enterprise applications can solve intelligent decisioning problem to a greater extent.
Data science offers a high degree of sophistication in making organisational and tactical decisions utilising data. Currently CIOs are charting out strategies for data science and advanced analytics across the enterprise. But many of the solutions are piecemeal and do not offer the degree of automation that guarantees repeatable efficiencies and seamless execution across the enterprise.
It is important that Data scientists create and design models which can be embedded in the workflow of business applications. They have to work with business practitioners to understand the larger context of the design and execution of the model. The biggest value proposition that packaged analytics brings in is the Objectivity & Efficiency in complex decision making and repeatable execution scenarios not to undermine other factors like reducing a wide range of inefficiencies around data preparation, model validation and more importantly human bias.
Embedding Intelligence in Enterprise Workflows
A few examples where ‘packaged’ analytics and intelligence creates efficient decision making in business workflows/applications:
- Wealth management advisory tools can have packaged analytics that profile and rank an investment product based on the context, life stage and risk appetite of investors. This can help a wealth management advisors garner insights about their customers but also reduce his own risk of under servicing with biased and non personalised advice based on his intuition.
- Small time stock trading accounts can be equipped with embedded analytics that can help in leveraging price & trade advantage to gain more returns.
- Digital customer engagement process can be enriched with analytical models that can help in discovering product affinity and churn propensity of customers.
- Chatbots and Virtual private Assistants can be enriched with Machine Learning and Advanced analytics bringing in a higher degree of relevance and cognitive element which can make conversation efficient and effective in improving conversions and customer satisfaction.
- Machine will not replace Human doctors in the near future, but EPR and health care applications equipped with advanced algorithms can offer suggestive approach to detect and diagnose life threatening ailments like cancer in early stages
- Retail Marketing could involve planning at a large scale which may involve customer value differentiation, understanding customer usage and adaptive behaviors, predicting inventory spikes and stock outs executed at Point of sale level. Such top down execution would greatly need packaged analytics to deal with contexts arising at store level to be rolled up to organizational level.
- Law enforcing agencies of governments can use crowd analytics added to video surveillance to manage and predict crowd movements in large events like congregations, fairs, etc. and managing and avoiding untoward incidents.
- The analytics under the hood of various popular apps like Amazon , Netflix driving recommendations and highly personalised experiences for their customers
In all the above examples, there is an incumbent application used by the enterprise for performing specific business function. The addition of packaged analytics improves not only the functioning and performance of Enterprise Applications, but also makes the user highly efficient in attaining his goals.
Challenges in ‘packaging’ intelligence
- Evolving a data supply chain: The strategy for success of analytics based execution involves nurtured investments not just on the infrastructure but also in creating a nimble supply chain of data. Right data need to be ingested, aggregated and enriched for a wide variety of downstream use cases in the context of enterprise and its current set of applications. It starts from the premise that organization might want to do it incrementally, than a big bang approach. Hence the overarching framework should be nimble and future ready. Systematically and progressively, at different junctures packaged analytics can be introduced and observed for improved efficiencies.
- Process for supervised intervention: It’s important that packaged analytics applications are designed as tools for aiding users, without re-engineering the existing process flows. Embedded or packaged analytics need to be implemented in a non-intrusive way to drive ease of use and enterprise adoption. The beginning phases of analytics adoption also need careful selection of training sets and treatment on outliers so that the algorithm learns the expected behavior to develop higher degree of accuracy.
- Trust, Transparency and buttoned control: It is important that in-house practitioners are taken into confidence about the relevance and use of packaged analytical models, which are embedded in the workflows of business applications. Specialist practitioners should be able to manage configurable aspects of the model to closely align with expectations in production environments. Impact should be validated with the quality of the output of the models and tweaked to drive outcomes as envisaged by business teams.
Enterprise applications with embedded intelligence presents enormous opportunities for digital enterprises to gain tangible economic benefits through better operational decision making and efficient execution. However implementing such a practice successfully will need a slight reconfiguration of how business teams and data science teams collaborate and the use of technology tools with right out of the box analytics and execution capabilities.