Intelligent automation is a popular subject nowadays. The technologies around which it is built – including artificial intelligence, robotic process automation, natural language processing (NLP), and cognitive computing – are frequent business topics.
However, these areas should not be seen in isolation from one another. They need to form part of a whole. What’s more, they need context – and that context is the very practical one of the organization within which they are to be implemented, and the processes to which they will be applied.
At Capgemini, this holistic view has been termed Intelligent Process Automation. In this piece, I’m going to look at two aspects of Intelligent Process Automation with which people are most likely to engage, involving technologies that enable human-computer interaction.
Voice assistants are perhaps the most obvious instance. They have two principal purposes.
The first and currently the most common application is to handle inbound calls. Their use reduces pressure on customer service teams, by enabling technology either to save some time, by asking for and obtaining standard information such as account details from customers; or to save a great deal of time, by handling entire customer interactions. Intelligence built into the system that recognizes words such as “I want to speak to an agent” can enable calls to be switched at any point to a customer service representative for exception handling.
This is what the baseline technology can achieve, but the full benefits can only be realized when this intelligent automation is built fully into processes. At Capgemini, we’ve been working hard to define and develop interfaces with major enterprise platforms such as SAP, AWS, and Salesforce to enable voice assistants to be tailored to individual corporate purposes.
The second principal application is the handling of outbound calls. In this application, AI voice assistants can contact customers to provide advice, for example, or contact delivery drivers to provide updates when they are en route.
As you might expect, a lot of development work is needed in this area. For instance, at a regulatory level in many countries, outbound voice technology needs inbuilt security to confirm the identity of the person answering the call; and at a practical level, we’re working to integrate outbound voice into enterprise systems such as MRP and supply chain management solutions to increase its applicability.
Overall, the efficacy of voice assistant technology and in particular, its intelligence, will be the result of continuous improvement. The more iterations there are – the more sample conversations and experience voice assistants can muster – the better able the technology will be not just at understanding the human voice, but at interpreting imperfect input, such as speech from which words are missing, or unclear, or seemingly out of context.
In this field in general as well as in specific areas including finance and healthcare, we’re working with the best practitioners in the market, working through typical customer scenarios, identifying pain points, and developing responses.
Text recognition – and beyond
Speech has a corollary in writing, not just as a means of communication, but in terms of human-computer interaction.
Just as voices vary in pitch, pace, tone, accent, and more, so handwriting can differ in a great number of ways, and for machines to understand it, optical character recognition (OCR) systems need to be integrated with smart algorithms and subjected to a great deal of practice before deep learning techniques can start to bear fruit.
It’s not just about handwriting, either. Unstructured data in general – which as well as handwriting includes video, audio, image data, and PDFs – is growing at a phenomenal rate. In 2017, IDC forecast that by 2021, at least 50% of global GDP will be digitized. That’s a huge amount of data, and Intelligent Process Automation will need to be put to work on much of it – for instance, reading and responding to a handwritten note from a customer who wishes to cancel a subscription, or assessing and processing scanned claims forms.
A case in point
Several of these intelligent human-computer interactions can be seen at work in a cognitive assistant we recently developed for a client. The aim was to increase the speed and lower the cost of a cash collection process, while maintaining the sense of human engagement – even though the function was now automated.
Our AI-based client solution contacts customers who owe money, making either courtesy calls or telling them payment is overdue. It has been implemented using cognitive NLP, voice transcription, the cloud, microservices, and various modern web frameworks, and it supports 24 languages, including less common ones such as Dutch and Finnish.
The solution has a semantic awareness of people’s responses, and notes their promises of payment. It provides a significantly reduced TCO in terms of both infrastructure and headcount, and because it’s cloud-based, it’s highly scalable.
All such technologies are of course still in development, and over time the benefits are sure not just to increase, but also broaden out and embrace new areas of our clients’ operations.
And that, of course, is the whole point for us of Intelligent Process Automation. It’s not just about what’s possible – it’s also about what’s practical, and directly relevant to our clients’ needs.
To learn more about how Capgemini’s Intelligent Process Automation offering can stimulate the erosion of organizational silos around your front, middle and back-office processes, resulting in the emergence of a new, borderless, highly automated client-centric organization, contact: email@example.com
Miroslaw Bartecki is head of Capgemini’s Intelligent Automation Lab focused on adopting AI technologies into business services. He leverages the potential hidden in deep and machine learning to increase the speed, accuracy, and automation of processes.
 IDC FutureScape: Worldwide IT Industry 2018 Predictions (IDC #US43171317, October 2017)