By Asheesh Mehra, Co-Founder & CEO, AntWorks
The value of the RPA (Robotic Process Automation ) market is expected to increase from $1.7bn last year to $2.3bn this year, with this figure predicted to more than double by the year 2022. One possible alternative to traditional RPA that can not only match its functionality but also offer more is Cognitive Machine Reading (CMR), an AI-based automation capability. CMR is able to read structured, unstructured, image and inferred data, whether it is printed or handwritten text, or simply image data such as notary stamps or signature verifications.
It can also undertake natural language modelling, processing and generation. For example, a system using cognitive recognition can understand that the phrase ‘patient’s name’ is the same as ‘Name of the Patient’ when processing documents, and as a result can manage that data far quicker than other tools.
Not only can CMR harvest structured and unstructured data, but it also processes the data efficiently, learning it in such a manner that it is able to automate entire business processes without human assistance and produce actionable key business insights. This makes CMR a tangible, all-in-one solution that can rival any other RPA tool and revolutionise the AI and automation industry.
Why are we talking about CMR all of a sudden, and why is it needed?
AI Industry on the verge of a shake-up
The AI and automation industry is on the verge of a major shakeup. Think of BlackBerry 12 years ago before the first iPhone was released. Apple changed the smartphone market forever – Blackberry didn’t see it coming and ultimately paid the price. The same is true of the AI and automation space right now as robotic process automation (RPA) vendors that use these technologies have fallen behind on their ability to deliver end-to-end business processing.
Although companies see the effectiveness of automating processes and the benefits that come with using RPA to analyse and process data, unfortunately they don’t always know when to start deployment or exactly what data that they should be analysing to yield maximum ROI. CMR on the other hand can structure data far more effectively than other forms of RPA.
Unstructured data: ignoring the obvious problem
AI and automation could be a $5tn industry by 2025, but it can only be as good as the data it runs on. Companies from a variety of sectors currently using RPA systems are unable to process crucial unstructured data, which includes anything from images, web pages, legal documents and medical records to mobile content. Rather, companies only benefit from analysing structured data in the form of standardised code text or categorised fixed field text – data that generally isn’t difficult to analyse or process, and only makes up a small percentage of overall business data.
In fact, unstructured data will make up 80% of enterprise data by 2025. If the real number is even close to that figure, businesses will never be able to use AI and automation to realise its full potential. This in itself is enough to question whether AI and machine learning technologies are being utilised to their full capability, and if not, whether these emerging technologies will even last beyond the next couple of years.
AI adoption: getting it right the first time
Most unstructured data is information that lacks any pre-defined data-model or properties and is usually difficult to analyse and process. As unstructured data makes up 80% of enterprise data, companies need to be able to identify the right type of AI tools to deal with it. These should be able to extract key data from both unstructured and structured documents. Currently, too many businesses are utilising RPA tools that are based on neural science. These can only analyse the structured form of data, where datasets are mostly processed one-by-one, and the system can’t recognise patterns in a particular dataset.
Fractal science-based AI, however, can recognise entire patterns and not just single characters, which means that it can sort entire sets of the same data at once rather than relying on human assistance or processing datasets one-by-one. For example, a system using fractal science can discern patterns on invoicing documents, ie. that the first three letters will be ‘INV’ and that it’ll have a series of currency figures listed on it, typically. As a result, this enables the AI to recognise and process a much broader range of information and characters.
True automation which harnesses CMR is the next big step. If you want to reap the proper benefits of AI for your business, RPA which is based on neural science, or that is not CMR enhanced, is not going to cut it when it comes to organising and making the best use of your unstructured data. Instead, focus your energies on finding solutions that are based on fractal science, and that have a clear CMR element. Only then will your company truly excel in the automation era.