Marc takes on the competition with ChatGPT


Pioneering is in our DNA. So, if there is a new technology with serious gamechanger potential, you can count on our KLM colleagues to dive on it immediately. This is also the case with generative AI, the revolutionary application capable of creating its own content - think ChatGPT. As an analytics translator in the Big Data team, Marc is always looking for ways to make the maintenance division Engineering & Maintenance (E&M) smarter. By developing new applications and preparing colleagues for a different work reality.

AI guide for the organisation

Marc is an analytics translator within E&M: ‘I fulfil a bridging function between the business and data specialists. Just like a product owner, but who mainly ensures that the right AI product is created. My role is slightly more strategic: I stimulate the use of these kinds of products. By collecting the needs of end users during the development process, but also by organising workshops and training sessions after delivery - and reporting on them to management. Selling the products, in other words. I don't just do this with colleagues, but also with customers, such as other airlines.'

When ChatGPT - the first version of generative AI that was widely accessible – first emerged, management asked our team to find out what it could do for KLM E&M. Exactly how I achieved this was up to me, I had a free hand. I was given a lot of trust and all the space I needed for my exploratory research. Eventually, I wrote a report on my findings. My conclusion was that there are endless possibilities, but for now there are four main application areas: coding, customer service, training, and technical documentation. Now it’s up to me and my colleagues to further investigate what those improvements could look like in specific terms.

Marc, analytics translator

Better looking back and forward

'At E&M, smart use of various AI applications can add a lot, which is why many projects are now running. The first one I’m involved in is AI-assisted labelling. When an aircraft part goes into maintenance, each of the thousands of possible situations gets a tag with a code: for example, event "LH Wing: upper side of wing several spots with missing paint" on a Boeing 787 gets the code 57-10. That way, we can keep track of exactly where maintenance is needed, but also which parts need repair more often than others and approximately when.

Reliable labels

So far, nothing new: the label system has been used within KLM for some time. But the codes have always been added by people - and that is where there is room for improvement. 'To help colleagues label, we have developed an AI application: our own version of ChatGPT, which is very user-friendly thanks to prompt engineering. At the back end, an extensive conversation takes place, but at the front end, a short description of the part and flagged problem is enough to put the chatbot to work. The AI starts searching, sometimes asking a few additional questions, and in this way determines the right code. This saves our colleagues a lot of time and makes the labelling process more reliable'.

Needle in a digital haystack? Found in no time!

Another AI application Marc is working on is Project TRAK: Transparent Reliable & Accessible Knowledge. This system should provide solutions in so-called AOG situations. 'That stands for Aircraft on Ground: if something is wrong with an aircraft, it will not fly. It's important to assess and solve this kind of situation as quickly as possible. To do this, our engineers need information from a variety of sources: from internal manuals to instructions from manufacturers and from files to old e-mails. In addition, we have a number of colleagues with an awful lot of knowledge in their heads - many of whom will soon be waving goodbye. It would be a shame if all that valuable experience were to be lost.'

'TRAK should secure and centralise knowledge about AOG situations. To do this, we bring the information sources together and make them accessible - and specifically, easier to search through. For example, if there a dent on the wing, somewhere it will say how big it can be before the aircraft has to be grounded (AOG). Thanks to TRAK, engineers don't have to sift through piles of manuals and bulletins. It's all in one place. In addition, new knowledge can be added, for example, using speech-to-text technology. Soon whatever problems engineers encounter with an aircraft on the ground, they will have the answer figured out in no time; knowing exactly what they have to do to get the aircraft back in the air as quickly as possible.'

Valuable discoveries

In both projects, Marc and his colleagues are working with state-of-the-art technologies. And perhaps more importantly, they have a direct impact on their colleagues' work. 'KLM encourages us enormously to keep learning. Testing new technologies and building our own applications, but also going to conferences and getting inspiration from others. Having the space to experiment and discover, but at the same time being able to make a difference in operationsfairly directly: fantastic, right?'

The crystal data ball of Sashiko and her colleagues

Predictive data helps with accurate inventory management of aircraft parts.

Woman looking at screen

Unique real estate variety at KLM Real Estate & Facility Contracting

Berend and Jeroen ensure that all our offices, hangars and warehouses are future proof.

KLM-office with two employees

The Repair Lab: where alternative repair solutions get to see the light of day

KLM's Repair Lab devises innovative ways to make aircraft repairs possible.

Man in worker space