Any sector where humans are completing mundane tasks is ripe for automation, including financial services, which over the years has included plenty of data entry and more than its share of paperwork.
The financial services industry sees this and is embracing robotic process automation (RPA), contributing to what is expected to be nearly a 12-fold increase in the market for RPA spending from $250 million in 2016 to $29 billion by 2021.
A host of options are included under the RPA banner, Dolphin Enterprise Solutions CTO Vishal Awasthi said, and those options grow and improve over time as the technology advances and understands how humans make decisions.
RPA can be separated into three classes, Mr. Awasthi explained. The first is based on the premise of giving the use tools to accelerate basic functions such as copying and pasting or jumping between screens. The next class integrates business applications such as workflow rules-based automation.
The newest level can process documents and images and needs a cognitive understanding so it can interpret natural language patterns. Systems can be trained to intercept incoming emails and extract attachments so people don’t even have to open emails.
“You can instruct the bot to do the extraction and content analysis automatically,” Mr. Awasthi said. “With it sitting in the cloud it can self-learn and you can share the fingerprint amongst all the applications that use that bot .”
Developers focus on making existing applications smarter by embedding such shared AI models in them without increasing IT infrastructure, he added.
Potential RPA advancements in bot technology are many, Mr. Awasthi said. They can be trained and retrained to understand new patterns or entirely new languages while getting to intent. Such supervised learning applications will allow RPA programs to scan the body of an email, or other natural language text data, to more accurately determine intent that was previous iterations could not extract. Whereas humans would previously have to review failed attempts and reassign them to the right categories, but now with machine learning, processes can be automated to correct and improve themselves.
Progress is also being made with systems’ capabilities with understanding multiple languages, Mr. Awasthi said. Bots can be trained to identify languages and automatically switch to the correct one or to offer assistance in the user’s first language. Programmers are also prioritizing emotion detection which can help prioritize queries and address disputes between employees or with suppliers.
“Sentiment analysis is a key aspect of natural language processing,” Mr. Awasthi said.