Our October article, The email challenge: easing the pain with artificial intelligence, highlighted a new AI project designed to automatically read and process unstructured emails and attachments in order to cut out manual time-consuming work. The aim is to use AI technology and natural language processing to extract information from unstructured formats and bring it into a structured form, so that it can be processed and digested more easily. So, what are the next steps and how far can this technology take us? And what are the wider benefits of this approach?
While the current attention for addressing the email challenge is focused on customer- or market-facing email exchange, the applicability of leveraging AI and natural language processing capabilities to read and understand unstructured information certainly applies across the functions and the organisation.
It is widely reported that 80% of all company data is unstructured, so the potential to move beyond emails is very clear. Companies have vast amounts of data that need to be analysed and processed. Much of that data is freeflow, and AI’s ability to process freeflow information, rather than just data in columns and rows, is technically mature and a step change from the past.
The ultimate aim is for AI technology to be accepted and seen as commonplace, says Gero Gunkel: “In terms of scalability, we can look at the example of Microsoft Word – it is everywhere and is just seen as a helpful business tool. For us, we want to be able to scale AI to every single user at Zurich, supporting all the business processes so that ultimately it is no longer seen as fancy new technology but is just a smart assistant that helps with the admin work, the work no one likes to do, acting as a helpful companion.”
The next evolutionary step could be driving towards proactive decision support capabilities, according to Armin Schaefer. “The next stage in the evolution could be taking the support to the next level, becoming an intelligent adviser, so not just bringing the information to the table but supplementing it with decision-relevant information. In simple terms, instead of giving an underwriter just the information that we have received from the broker, it could also be enriching it, either from our historical data or external data sources, drawing conclusions and proactively supporting the underwriting process.”
AI technology is not about replacing humans but supporting and complementing them, says Mr Gunkel: “In terms of decision-making support, I am a strong believer in human ‘gut feeling’ that a machine cannot copy. But what machines can do is read much more than humans. Machines lack high-order intellectual traits such as reflection, wisdom and curiosity. So while I don’t think that in commercial insurance we will yet have a machine that can decide pricing, there could be a machine that can read 10,000 submissions and inform the underwriter, at the point of underwriting, about a certain pattern it has found, highlighting anything that is unusual or peculiar. AI is that dependable worker that reads everything in record time and never gets tired and never forgets. This can be a very good complement to an underwriter that cannot memorise everything but has a very good intuition about things – a match made in heaven.”
Far-reaching and lasting benefits
One obvious and universally applicable benefit of taking an intelligent approach to the email challenge is speed: by automating the processing of emails, the insurer is able to respond faster to clients and brokers, whether it be enquiries, submissions, claims or policy changes.
And as Tobias Wild points out, it is not just about speed of response, but also quality: “We have found that when the machine is processing data, there are less errors and there is a more consistent approach and procedure on how to extract data than you would have with a team of individual humans, who may all have their own approach to this.”
By bringing structure into the freeform flow, we are able to allow downstream/back-office system integration, remediate rekeying and drive automation. In addition, there is better processing transparency, which enables analytics and supports process re-engineering and optimisation potential. It can enable people to have a much more data-driven quantitative conversation about their business processes, for example as to why there are certain bottlenecks. Analogue blackspots can be avoided and management KPIs can be enabled.
There is also the issue of capacity-generation – the tool is available 24/7 and brings a lot of resilience into the game, so that teams can refocus away from mundane administration work to valuable underwriting or claims work, focusing on delivering quality customer service from the moment they enter the office.
The time is right for AI to play an important support role and it is already proving its capabilities and value, but we are still only at the beginning. The outlook, from a technology capability point of view, is extremely promising. And ultimately, there is a vast playing field of opportunities for AI to add value – a wide-open horizon.
Contributed by experts from Zurich Insurance Group: Armin Schaefer, head of digital and new technology, commercial insurance international programmes; Gero Gunkel, chief operating officer, data science leader; and Tobias Wild, lead digital business analyst at digital and new technology, commercial insurance international programme business