The following blog post was written by Avanade alum Krishnan Rajaram.

A large majority of clients are looking to leverage robotic process automation (RPA) and artificial intelligence (AI) technologies but often struggle to select the right use cases to start the journey. Internal discussions to prioritise the use cases often run into months as executives debate the business case, people impact, security considerations, data availability and the like.

Companies often need to strike a fine balance between speed and impact, i.e. select use cases that are impactful and will generate the right internal pressure while ensuring results are visible in weeks or months, and not years.

One of the ways of breaking this stalemate is to look for areas in the business where people are “thrown at a problem”. At times, this happens as a reaction to a regulatory change that requires a timebound response but there may be other business triggers as well.

Below are two examples where RPA and AI technologies have helped companies remediate a situation where an army of people had been ramped up to deal with a high-volume, repetitive, timebound activity, and it became evident that it was neither effective nor sustainable.

The first one is from the Utilities industry, while the other example is from Financial Services focused on Anti Money Laundering (AML) compliance.

Reducing costs for a growing number of utility customers

Ofgem introduced the Cheapest Tariff Message (CTM) regulations in 2014 requiring energy suppliers to provide customers with information about the cheapest tariffs available and how much they could save by switching to them.

For many energy providers, this meant wading through multiple legacy systems and data silos to compare the energy consumption of each customer against the myriad of complex tariff plans and determine if a customer could benefit from a cheaper tariff.

Some providers employed armies of people to perform this analysis each month. Some were even contemplating if paying the fine would end up being more cost effective!

One of the energy providers found that despite employing many individuals, they were not able to achieve 100% coverage of their customer base. When it became clear that there was no way they could keep growing that team, they decided to use the opportunity to try RPA as a possible solution.

The initial scepticism from parts of the organisation, especially given the regulatory angle, was overridden by the top executives who were convinced they had to find non-linear ways of solving the issue.

The pilot provided highly effective as the process itself was relatively straightforward but the complexity of legacy systems, data silos and high volumes made it time consuming and expensive to execute manually. This was a BIG internal win for the company given the high visibility. This is an area where RPA solutions can work wonders by mimicking the steps a user would take at the user interface (UI) level.

Identifying potential security threats in the banking environment

Banks are obliged to monitor every transaction with a monetary value and determine whether the transaction is suspicious. Such transactions are communicated to regulators via Suspicious Activity Reports (SAR).

Traditionally, banks have used software solutions plus armies of people to spot suspicious activity. However, by setting strict thresholds on transactions, they have created a different problem, i.e. “false positives”. False positive transactions, at first glance, seem to have the characteristics to trigger an AML alert but are bona fide transactions citizens or companies.

Large banks generate tens of thousands of false positives per day, which needs to be investigated. This can be a highly manual activity with low yield and often impact to customer service. Meanwhile, regulators are not pleased with being dumped with large volumes of false positives while criminal transactions flow through anyway

This is an area where Machine Learning (ML) models combined with RPA technologies can help spot suspicious transactions, reduce the volume of false positives and reduce backlog. As machines learn from the vast volume of data, we find their ability to detect patterns and predict suspicious transactions also improves.

If we look deeper, each business has pockets like this where people have been thrown at a problem until a long-term solution is identified. As priorities evolve over time, these teams continue or even grow and become ideal candidates for AI.

Are there areas in your business that fit these criteria?