Friday, June 21, 2024

EXCLUSIVE: ‘A Greater Understanding’ – Daniele Cordioli, & Chris Pearce, esure in ‘The Insurtech Magazine’

Anyone familiar with the work of comedian Steve Coogan and the antics of his alter ego Alan Partridge, may recall a scene where the hapless DJ tries to order a cinema ticket on the phone, via a speech recognition bot.

It’s 90 seconds of delicious torture, a she repeats the word ‘Inception’ over and over again, using different intonation and inflection with increasing levels of desperation and, ultimately, despair.

It’s only when you start breaking down what computers need to be programmed with – or machine-learn – that you realise just how awesome our brains are when it comes to understanding the complexities of language, its context, nuance, sentiment and syntax – and then responding appropriately.

Natural language understanding (NLU)is all about eliminating those elements of the written and spoken word that are lost in translation between us and ‘them’, thereby ensuring that the right outcome is achieved through natural language processing (NLP).

It’s a branch of AI that has actually been around since the beginning of early computing and it’s said that, back in the 1950s, a translation from Russian to English was the first example of a natural language application. It’s an apocryphal tale: a computer was asked to translate the biblical saying ‘the spirit is willing but the flesh is weak’ into Russian, and then back into English where it reappeared as ‘the vodka is good but the meat is rotten’.

It was clearly early days… but it does illustrate the crucial role of NLU in recognising not just what has been said or written, but, more importantly, the intention behind it.

A lot of water has flowed under the digital bridge between human and machine cognition since the 50s, of course, and machine-learned nuance and context inform exchanges in so much of our everyday lives and over multiple channels today.

Developments in NLU allow us to have what feel more like proper conversations with virtual assistants like Alexa, Siri and Cortana and, as a result, they know, for instance, that we probably want a local weather forecast when we ask ‘Alexa, what’s it like outside?’, rather than a description of the street on which we live or indeed an existential summary of geopolitical events.

Understanding is key to progress of any sort – and it’s fundamental to how we leverage the sheer superior lifting power of computers when we want to do things faster, more accurately and more profitably.

Intelligent at the core

The case for broader AI adoption in insurance had already been made before the events of the past two years brought any laggards abruptly face-to-face with it.

In 2019, Lexis Nexis’ The State of Artificial Intelligence And Machine Learning In The Insurance Industry report, suggested those that had adopted AI and machine learning witnessed 88 per cent faster claims settlement, 88 per cent better cross-selling, 87 per cent better fraud detection, 85 per cent better risk scoring, 80 per cent improved pricing decisions, and 78 per cent improved profitability, when compared to those that had not. Three years on, insurance businesses must be brave enough to rethink their structure and model, rather than just tinkering at the edges with AI, says Daniele Cordioli, head of solutions consulting EMEA for, a specialist in NLU. They should put it at the core of their operations.

The challenge for data scientists, meanwhile, is to accelerate the sophistication of NLU so that they can confidently apply AI across organisations to address three key issues for insurers, which Cordioli identifies as: reducing processing time, eliminating bias and improving risk and pricing, all of which have positive outcomes for the industry, the company and the individual policy holder.

“Insurance is at least five-to-10 years behind [other financial services], in terms of technology adoption and innovation, ”says Cordioli. “But the pandemic dramatically changed perception and awareness. The carriers realised, during lockdown, that they couldn’t properly sell, they couldn’t sense the risk. In other words, they couldn’t guarantee business continuity. So, something had to change– and, thankfully, it is.

Chris Pearce, head of data science at esure Group, a UK-based personal lines insurer, agrees: “You’ll see the vast majority of companies now aiming towards digital models – self-serve, multiple touchpoints, applications – forming a core part of the strategy and technology. Data and AI is at the heart of these ambitions.”

Insurance is undoubtedly data-rich, possibly with more access than any other branch of financial services to granular and very personal information about customers: their health status, driving habits and attitudes to high-risk pursuits, for example.

Data is food for AI, and the more it can gorge on, the more likely it is that technology of this kind can make a real difference – and not just in customer-facing processes (those chatbots that so befuddled Alan Partridge) but also in back-office applications that are of more importance to insurers than ever before. Just a couple of months into the pandemic in 2020, a Celent report, Property Casualty Insurers Weighing In On Emerging Technologies, showed 67 percent of large property and casualty insurers had identified cost reduction and process improvement as key business targets, and those challenges haven’t gone away– but they can be addressed with NLU and NLP. Such sequential solutions can apply structure and interpretation to complex, language-based documents in the claims process, for example, to reduce underwriting leakage and substantially increase the accuracy of a policy review.

“I can identify tangible benefits [for using AI underpinned by NLU] in better risk assessment, for example,” says Cordioli.“There is also better knowledge of the customer base and behaviour, which impacts customer engagement and claim management, too.

“The more data you have, the better you can profile the customer and extract a lot of relevant knowledge. You also get better awareness of the risk exposure. In the last year, we’ve found a lot of interest in the space of cyber policy review, for example. “Insurers’ overall objective appears to be shifting from a detect-and-repay model, to a prevent-and-predict one.

”So, there are big plusses for insurers using this branch of data science. But what about customers? They continually provide unstructured data via multiple touchpoints, so how does AI transform this into something of benefit?

Pearce says: “AI can uncover the minutiae of details and interactions we have, which can often be missed by the human ear or eye, help refine process proposition and create new ways of dealing with, prioritising and supporting the myriad of needs our customers have.”

Cordioli continues: “The policy holder will see a quicker processing time for claims submissions, for example.

“Having more data, being able to process this data, and extract more knowledge from it, allows insurers to be more reactive in their engagement with the customer.

“Another benefit is objectivity. Artificial intelligence can remove some of the bias that is part of being human. We can introduce this technology across the entire value chain, and expect it to work in the same way, with less subjectivity.

“The third benefit is a better assessment of the analysis we are doing, around risk, for example, and being able to link risk and pricing in a better way, so there is tangible benefit for policyholders.”

But, he stresses, artificially-intelligent technology will achieve its maximum potential only if it’s seen as a catalyst for deeper change within the organisation.

“It’s not a case of plug and play – put AI into a company and, like magic, everything works. You have to create a culture around it; a transformation has to start.”

Pearce agrees. “AI is not just about predictive modelling, chatbots and recommendation systems,” he says. “It’s about a company mindset, a cultural change, and a unified approach across the business. It’s about bringing cross-functional collaboration, human creativity and hard science together, harmoniously and robustly, in order to understand and meet the needs of the people you service. All of that often adds up to transformation. But, if you do that right, then AI becomes the driving engine of decision-making.”

“Technology and data are just two of the four pillars that are fundamental to successfully delivering a product; we also take into consideration people and process,” says Cordioli. “By people, I am not only referring to data scientists and everyone working in the ecosystem of AI, but also stakeholders and people in the business units who have to embrace the process and methodology.”

Assuming that is successful, how will these tools affect the industry? “Never has this scale of change at once represented an opportunity to radically alter brand, mission, the way the market works and the products on offer,” believes Pearce. While Cordioli says becoming more technology driven will ‘probably be the only way to compete in the market in the next 10 years’. “The cool term ‘insurtech’ will probably disappear, because all insurers will be insurtechs,” he says.


This article was published in The Insurtech Magazine #07, Page 37-38

  1. Peach Announces Pipe Partnership to Power Working Capital Solutions Read more
  2. TBC Bank Uzbekistan Secures $10 Million Loan From responsAbility Investments AG Read more
  3. Invessed Set to Help Wealth Managers Respond to Disengaged Demographic Read more
  4. Singapore Publishes Updated Money Laundering National Risk Assessment Read more
  5. Why Age Matters: Mads Pedersen’s Fresh Take on Venture Capital Read more