Intelligent automation leaders are under pressure to deliver impactful, scalable and continuous improvement. Yet their toolbox of solutions contains a major blindspot - manual email processing. Fortunately, natural language processing has come of age. It addresses the weaknesses of automation, analysis and digitisation tools, completing the tech stack for intelligent automation.
Continuous improvement requires many different factors. It needs visibility into systems to identify waste and breakdown, the ability to track and monitor processes in real time, and straight-through processing for ease and efficiency. In short, it needs intelligent automation.
It’s well-established that a combination of different automation and analytics technologies is required to achieve this. In core business functions, this tech stack typically includes Process and Task Mining for measuring process performance, Optical Character Recognition (OCR) and Intelligent Document Processing (IDP) for digitisation, and Robotic Process Automation (RPA).
However, the intelligent automation stack is incomplete. Even with all these tools, change initiatives all too often fail to achieve meaningful, continuous change. Nearly three-quarters (72%) of shared services organisations have implemented RPA, but the largest cohort (46%) achieve less than 10% in savings. This is because current approaches have no answer to the manual communications processes - namely email - that comprise most work in shared services.
NLP: The missing piece
Every part of the current IA stack plays a critical role in achieving continuous improvement. Yet every tool depends on clean, structured data. It’s the lifeblood of intelligent automation. Process and Task Mining are blind without it. RPA can’t process what it can’t understand. The emergence of OCR and IDP lies in this need for digitisation, creating structured data from physical and digital documents.
However, shared services run on unstructured communications. The bulk of an agent’s time is spent performing routine, email-based tasks - reading a service request, triaging an email, and following up with internal customers for more information. While these tasks are low-skilled and repetitive, they can’t easily be analysed or automated due to a dependence on freeform natural language.
As crucial as current intelligent automation tools are, they aren’t truly intelligent. We’re a long way off artificial general intelligence. We still rely on humans for contextual awareness, understanding language and subtext. Consequently, agents are stuck in email while intelligent automation leaders have little choice but to look for process improvements elsewhere. Yet, as there’s no understanding of the most inefficient comms-based processes, this usually results in only minor changes and diminishing returns. This isn’t surprising when 43% of shared service functions struggle with process technical complexity, and 41% with siloed automations.
This is why interest is building around natural language processing (NLP). Recent technological advances mean that NLP can accurately process and understand masses of unstructured communications data at speed and at scale. Its ability to predict word meanings and understand the intent behind conversations is also constantly improving. Transformer-based NLP models now even routinely outperform humans in advanced reading comprehension tests. Furthermore, a recent focus on low-code and no-code development means that powerful NLP models can be trained up and deployed faster than ever.
This is why 67% of intelligent automation teams say ‘data extraction from emails using NLP’ is an important investment priority. NLP is poised to bring visibility to email communications, structuring unstructured information, identifying the points of process failure and highlighting fresh automation opportunities.
Completing the stack for Intelligent Automation
NLP has a recent but increasingly important role to play in the intelligent automation stack. It serves as a guide and enabler, uncovering the areas of process breakdown and failure for deeper analysis and automation.
Process and Task Mining are only one side of the coin for process improvement. The other side is Communications Mining, an application of Conversational Data Intelligence enabled by NLP. This is the analysis of unstructured communications data to find the root causes of process failure. Process Mining shows you where a process has broken down, but Communications Mining explains why. It forms the first step, helping change leaders identify problem areas for deeper analysis by process and task mining.
NLP is also a key ally for OCR and IDP. Working together, these tools have the potential to digitise all information in the enterprise. NLP structures all email and comms-based processes, while OCR and IDP digitise email attachments, images and documents. For the first time, intelligent automation leaders have clarity and visibility over every part of the business.
Finally, the structured data created by NLP feeds into and enhances RPA efforts. Manual email processing becomes a structured process, opening it to automation as well as analysis. For example, an RPA software bot can be designed to automatically reply to common requests, using NLP to understand the intent and next-best action. Agents can confidently hand off much of their workload to machines, boosting productivity and leaving them to focus on value-add.
Yet, NLP not only complements the existing tech stack for intelligent automation - it completes it. With clarity and visibility into the source of all workflows, change leaders achieve true end-to-end process improvement and automation. There’s no longer any break when a process enters an unstructured funnel, no gap in intelligent automation. Transformation leaders can touch and improve the whole process. Cure the cause rather than treat the symptoms of process inefficiency.
