A Practitioner’s Guide: Algorithmic underwriting in commercial and specialty insurance

The report, which is available to download for free, examines where industry adoption is right now, and focuses on three areas:

• Clarity – A clear understanding of what algorithms are and what they do
• Understanding – A better appreciation of what Artificial Intelligence (AI) and machine learning are
• Guidance – Practical steps for implementing algorithms and machine learning in insurance

Introduction by Robin Merttens, InsTech London Partner

There is a kind of mystique to algorithms when used in underwriting. Some think of them like magic dust that is sprinkled on data and immediately intelligent underwriting occurs.

One day, algorithms will transform how we do commercial insurance, but right now, most underwriting algorithms are quite simple.

Consider autonomous cars. They are driven by intelligent algorithms that have been in development for decades, but it will be many more years before they fulfil their potential. So it is with algorithms in commercial insurance.

Building algorithms is not that difficult – not much more than running some data through some rules to get a decision. As we’ve said before in previous reports, the technology to build insurance algorithms already exists and is reasonably commoditised, but there are two significant barriers slowing wider adoption:

1. Mastering the data – Sorting out data so that it is of a quality, format and accessibility to feed an algorithm on demand.
2. Access to the talent and skills – The relevant skills to implement scalable frameworks for automating underwriting decisions rarely reside within the standard insurer skill set. They either need to be acquired or outsourced.

Algorithms that automate the pricing of retail insurance lines are now mainstream, and the use of bots to deal with customer queries is seeing rapid adoption too.

While there are lessons to be learned from retail, the commercial and specialty markets have different challenges because the risks are more complex, less standardised and there are far fewer transactions meaning a smaller number of patterns to learn from.

Our goal is that having read this report, you will have a clearer understanding of what the terminology means, how algorithms will change underwriting, how companies are starting to implement them and what lies ahead.

Navigating algorithmic underwriting in Property & Casualty – InsTech London Podcast 160

Robin was joined by Andrew Yeoman, CEO and Co-founder of Concirrus, on Podcast 160 to discuss the findings from the report.

The conversation covers what’s needed to move the technology forward, plus practical examples of how Concirrus is using algorithms to identify marine risks and build dynamic pricing models.

It’s available to listen to on our Podcast 160 page, and through Apple Podcasts, Spotify and Google Podcasts.