BMW April 2022

Here’s the classic finance-led vision of the future: Electric vehicles, autonomy, block chain. It’s almost as lazy as the average finance article which takes seconds to name-check the Silicon Valley banksters. Will this touch our lives? Yes. However, if we are aware of what this is, we can turn the push for ‘Big Brother’ from domination by remote foreign investors into something much more useful, and profitable too.

Connectivity and over air updates

We are all familiar with our phones and computers receiving new software updates and alerting us to imminent installation. In part the frequency of the updates is related to mitigating cyber security risks, and although inconvenient it does not matter if the device ‘freezes’ for a few minutes.

That logic has now extended to cars

Gone are the days of going to a dealer for software updates. Now the vehicle configuration, software improvements and even performance upgrades can be delivered directly to our vehicle. The UK parc still has the majority of the vehicles that are not enabled to do this, but most new vehicles on sale now can do this. If you are not the vehicle owner or the vehicle manufacturer/approved repairer, can you see this? No.

It generates another question. If the vehicle was built with software pack ‘V1.00’, delivered when ‘V1.01’ became available and had an incident some years later by which time ‘V3.20’ is installed – what is the insurance company covering? Even if the vehicle user knew what all of this meant, how can this be communicated with service providers such as insurers? 

Now consider this. A fault on a vehicle can be altered by software, but a vehicle electric system usually has a series of switches and actuators. If the these fail or malfunction, there is a limit to what ‘over-air’ updates can achieve. That applies to pure electric vehicles as much as internal combustion powered vehicles. So, the idea of no physical repair is somewhat ahead of reality, by quite a few decades. Humans 1, ‘Big Brother’ 0. 

Block chain technology 

This takes some decoding, but in essence this is a means of sending a package of information from one place to another, swiftly and securely with multiple audit points to ensure it does not get trapped in a false destination. So, if we think about money, such a system is more likely to resist fraud than most. It is the core technology of encrypted currency. 

A computer performs by calculating millions of times, creating little bits of answer to build towards the ‘correct’ answer, whereas we can perform the same calculation in fewer, bigger steps. The computer can perform these mundane small-scale calculations so quickly and accurately as to appear better than a human. Conventional computer calculations are in essence ‘dumb’ in that there is no imagination, no alternative view except the very obvious path.

If we add vehicle damage parts lists images or processes to the financial information, we can move the package from place to place with a full point by point audit – which is very powerful. Humans ‘0’, ‘Big Brother’ 1.

Artificial intelligence (AI)

Until now Block chain technology has been primarily used for financial transactions. The addition of an ability to compare slightly different results with a ‘new situation’, and then calculate how the ‘new’ situation compares, is dubbed Artificial Intelligence (AI). It elevates a computer from being dumb to slightly less dumb, where programmes are written to enable the machine to appear to take decisions. These ‘decisions’ are based in programmed rules.

AI has arrived in the collision repair business, with an overnight success that’s been a decade in the making. AI has been through a range of applications, from ‘portal scanning frames’ to high-end 3-dimensional camera technology, to end up with images uploaded from a mobile phone. That’s because mobile phone cameras have migrated from a simple 2-dimensional frame capture device to a video live stream with the ability to take static images. Along the way the promise of untold wealth has burnt a few people, including Mr Aquilla whilst he was CEO at Audatex. 

Take a dent on a door skin. Using a relatively low-tech camera on a phone, driven by the user who may not be able to get the best results, the software application guides the user to take images in a specific order. The software is then able to compare those images with previously recorded images, which then identifies:

  • Make/model
  • Panel location
  • Type of damage – scratch to dent.

From this the software calculates the job, ranging from the paint as well as blend area, through to ‘not for me’. Each time the system finds a variation on what it knew, the ‘knowledge’ base builds. Fundamentally an AI system is in essence dumb, but expert at classifying things once it has been programmed to expect those things. Humans 0, ‘Big Brother’ 1.

 AI and structural damage

For the time being such systems dominate the light scratch and scuff repair market, where estimators are rarely used. Instead, the client is given an indication if the job can be done and if so, what the cost will be. When a repairer from the company turns up to see the vehicle, they can qualify the repair and advise on additional repairs. 

There are of course estimates for more serious collision repair which also have a triage from the same type of low-quality images – and that’s where things get tough. Currently most ‘AL’ applications are focused on bumpers, fenders and doors, but not on dimensions such as the wheelbase. 

You see a vehicle damage assessor will start to think about the primary direction of impact, where the impact energy went and to consider if there is evidence secondary impact. So, if we have driven over a parking place kerbstone at 15 km/h (have you ever seen my driving?) then it would be reasonable to expect the underbody is going to get some sort of damage. At 30 km/h that damage would be much more severe. 

If we add vision analysis to such an ‘AI’ system – remembering the mobile phone will probably have video streaming and so can analyse snap shots just microseconds apart to calculate distance – the system can be taught basic parameters in addition to what is already known. Suddenly measuring the wheelbase from one side to the other on the target vehicle when knowing there has been a heavy offset frontal collision, confirms many things. Many things which would prompt a human to consider the consequences, whereas AI, for the moment, struggles:

  • Has the chassis rail bent?
  • Has the subframe moved?
  • Has the damage extended to the upper A Pillar?
  • What systems in the locality have been crushed or damaged?

Humans can easily visualise what has gone on beneath the skin, and at least raise concerns pending verification once the repair is underway. Humans 1, ‘Big Brother‘ 0.

The key message

A human vehicle damage estimator should be able to accurately scope what has been damaged and raise concerns for further investigation even when using only photographs. From a mix of research, knowledge and experience the estimate will be produced in a faster way than present AI systems. However, AI is not standing still, and there is quite a push to eliminate human estimators.

For the next decade the best outcome is to have AI enhance the human vehicle damage assessor’s work, to ensure no aspect is forgotten. This would ensure better consistency and allow humans to do what is best – thought, imagination, verification, composing – in the pursuit of the perfect vehicle damage estimate, faster than ever before. Humans 1, ‘Big Brother’ 1; win-win.   

Auto Industry Consulting is an independent provider of technical information to the global collision repair industry via EziMethods, our online collision repair methods system. 

For more information please visit the website:

By Andrew Marsh