Car Dealer Magazine has reported how CapHPI owner Solera is cutting up to 50% of CapHPI’s workforce to be replaced by AI. Fleet World has reported that some 50% of staff could potentially leave the business, replaced by machines. CapHPI PRs have attempted to soften the impact on valuation subscribers by assuring them that there will still be some human oversight, but with half the humans said to be leaving the firm who knows how true that might be.
CAP HPI Valuations
Cap HPI was formed following Solera’s acquisition of Cap Automotive for £295.5m in 2014, bringing it together with vehicle history specialist HPI, which the American firm had purchased in 2008 and the two businesses were subsequently combined in 2015, so they have been working together for over 10 years.
The news is not hugely surprising. I would be doubly surprised if the headcount from the 2015 merger had not already been reduced by a significant percentage. Valuation companies have used AI via mathematical models and algorithms for well over 20 years. Does a subscriber still remain who believes that valuation companies are not already using AI to analyse market observations, pulled from auctions and advertising platforms?
My experience over recent years spent arguing on behalf of policyholders in insurance write-off disputes shows just how wrong AI models get valuations. I spend a lot of time disputing Guide numbers, including those of CapHPI. Much of the debate is where older cars overlap or outrank certain versions of newer models.
The car market is not always intuitive, and the way depreciation curves change as vehicles age is never straightforward. There are so many errors made in mathematical modelling of depreciation curves, and how vehicles lose or gain value over time. An AI model will rarely weight characteristics and market evidence correctly, especially as vehicles age.
CapHPI, of course, is not alone in this shift to machines. All valuation firms will move here, because any firm’s primary aim is to make money for shareholders and humans increase overheads. Accuracy of data and integrity of modelling does not sit at the top of the list of priorities.
I see this daily in insurers using Guide data – from all providers, not just CAP – as a bulletproof vest that they can wrap themselves up in and say, “Well, it’s not us, it’s them; this is what they say and what our insurance contract says we will pay out. It’s not our fault if the data is wrong.” Weaker accuracy across an entire market model is what I have come to expect from valuation companies, particularly at the older end. With the added deterrent of high friction employed as an integral part of the claims process, many policyholders just accept the below-market valuations and vow never to use that insurer again.
Defending independent valuations drawn from adjusted market observations against a tranche of incorrect AI valuations, where Guides are essentially valuing off each other (an historic anomaly within the industry) and novice Ombudsman adjudicators with little experience in recognising and discarding incorrect Guide numbers is one of the big challenges that I face in my work on a daily basis.
As I see it, further shifts towards AI in mass market are inevitable, and increased numbers of disputed write off valuations are also inevitable as a direct consequence. Cuts to any professional service which addresses an important consumer need – such as valuation providers and insurers – usually hurts someone else down the line.