Having delved into the worlds of perception and cognition in AI, it’s time to turn our attention to the third and final pillar in the pursuit of artificial intelligence – action.
Once AI has gathered and processed data about its environment it must act to maximise it chances of success in carrying out a specific goal. That goal can be anything from winning a game of GO to making a profit on the stock market to safely steering a car down a dual carriageway without disastrous or expensive consequences. But what’s important in fulfilling all of these tasks is that the system automatically affects its environment in some way, whether that is by manipulating software or physical objects themselves.
Robotics has been around for some time and use cases in the manufacturing industry are well documented. This tried and tested technology is now being married with Artificial Intelligence, so that instead of performing pre-programmed repetitive tasks robots can flex their actions based on changing environment and goals. These reactive robots are being used in a whole host of ways from caring for the elderly to self-flying drones.
As much as I would love a mechanical AI robot to help me out around the office (a 3pm head massage would be ideal); this type of robotics has less relevance to most organisations than manipulation of software – so-called ‘software bots’ or simply ‘bots’.
Robotic Process Automation (RPA) is a hot topic right now, across the finance sector in particular. Software bots manipulate multiple screens to automate repetitive back-office processes such as data entry. Initially, these tools had to be explicitly programmed to perform set processes and couldn’t handle any deviation from the expected norms.
Enter AI. If you upskill your RPA bot with the ability to gather data and learn, you can automate a plethora of complex and variable tasks and processes. This intelligent automation learns and adapts as it goes and can automate entire end-to-end processes and workflows.
Take insurance claim processing. Ipsoft’s bot Amelia can absorb claims rules, engage customers directly to provide required information and use this data to automatically manipulate multiple systems to process a claim. And she never needs to take a break to grab a coffee or have a 3pm head massage. Japanese insurance company Fukoku Mutual Life Insurance hit the headlines recently when they laid off 34 employees and replaced them with an AI-enabled RPA system based on IBM’s Watson, which can handle financial pay-outs: tough news for both the outgoing staff and their local Starbucks.
But it’s not just insurance claim processing that is being automated by these tools. Work fusion’s virtual data scientist combines machine learning and process automation to automate a whole host of tasks. It learns the ropes of a complex process from historical data and real-time human actions. It then uses this information to train machine learning models which automate human judgement. So, as it can extract and categorise unstructured information with greater accuracy that a human, it can be used to automate processes that normally require human experience and judgement, which many see as the Holy Grail of AI.
Al businesses have manual, non-revenue generating processes that take up valuable staff time. The problem is acute in organisations who have legacy systems which tend to require more manual effort to get data in and out of them. The use of artificial intelligence to automatically manipulate existing systems, even legacy ones, has obvious appeal and advantages, and is an area that could be set to dominate the growth and development plans of cross-industry sectors for decades to come.
