Case study: How FORVIA Faurecia is using AI to reduce insurance premiums

This year represents the first time we’ve applied artificial intelligence to our property insurance process to calculate our insured assets – with great results.

We’ve saved thousands in premiums, reduced time and increased efficiencies, all while producing accurate calculations.

Calculating insured values is a specialist, complex and time-consuming task – particularly for an automotive supplier such as FORVIA Faurecia, which equips one in every two vehicles globally with its products on average.

Our operation is complex, with multiple assets across multiple locations around the world. To make these calculations requires about 80 controllers, who conduct countless audit and site visits across the world; and who examine assets that are constantly evolving – either expanding or reducing in size and scope.

We were paying out more in premiums than we had to because estimations were inaccurate. They were simply that – estimations.

Where there were changes to an asset, an amortised value rather than a replacement value was provided. This meant that we had to conduct several reviews and make iterations to the proposed values.

Add to this the complexity of applying inflation to estimated values, and often this meant paying much more in premiums than we would otherwise have to.

For example, applying an average of 10% inflation rate also meant applying 10% to the value of our properties and 10% on our insurance premiums. Precision was difficult, if not impossible.

ENTER AI

That is, of course, until we implemented AI into our property insurance process. This brought us much closer to accuracy.

The model calibrates the best values and appraisals based on an expansive database of the values of equipment, machinery, buildings and other specialist assets. Algorithms assess our assets against [what’s] in the database to quantify our values as close to reality as possible.

The impact to both our cost and operation was significant. Before AI was applied, it took our controllers about six months to complete the assessments and produce a report of our insured assets and values. AI cut down this process by half, now taking us only three months to complete.

Where our 80 controllers would take day trips to visit multiple sites, they can now conduct assessments from their desks. It also means that we do not need to involve as many controllers because a lot of the process is automated – again reducing time and increasing efficiency.

Perhaps one of the most impressive results is in how AI helped in terms of inflation. So rather than applying a blanket 10% inflation rate based on the inflation per country, we were able to calculate using more specific indexation based on equipment.

This meant that, in some cases, the rate was only 2%. With greater accuracy, we can drive down the cost of our premiums.

MEETING RESISTANCE

We took several steps to make this happen and we encountered many challenges along the way. The first step was finding a suitable AI model.

Luckily for us, this came relatively easy as the concept of embedding AI into our insurance process came from the company we use for site appraisals.

I would recommend to anyone looking for an AI supplier to consider the propositions available within your network of existing suppliers – leverage these relationships.

Next, we needed to convince our lead and co-insurers to accept the AI-driven calculations. We expected and prepared for some resistance and challenge. After all, we were proposing a new process that would potentially lead to paying lower premiums.

The key for us was in being overtly transparent about all calculations including the method behind how these calculations were being made.

We experienced a lot of disagreement, for example, about the calculations of our exposed assets. The main pushback was that our insurers considered this as ‘more known’ before the new process.

Our response was to demonstrate how the values were calculated. We had upfront and honest conversations involving local plants, our AI supplier and our insurer.

In the end, collaboratively, we finetuned the values and were providing an assessment that was consistent. We also applied this level of transparency to obtain buy-in internally.

As well as an increase in value, the algorithmic appraisal demonstrated a decrease in value in some of our assets and we were worried about how that would be received by management. However, this level of transparency worked in our favour.

NEXT STOP? THE PLANET

So, what next for us and AI?

Other than making this a routine part of our property insurance process we will also be applying AI technology to our risk assessment of climate-related exposures.

AI has given us a vision. We can foresee a nearfuture of more accurate diagnostics even with complex risks such as climate change and we aim to use the data available to adapt our process.

And I hope to report similar positive results with our climate goals as I have done with our property insurance.

Michel Josset is director of insurance and loss control at FORVIA Faurecia and a member of the AMRAE board of directors.