Fast 5: IFS Pursues Aftermarket Artificial Intelligence Strategy
![Rob Mather](/sites/default/files/styles/crop_freeform/public/2024-08/rob-mather-ifs.jpg?itok=7v6CmYl3)
Rob Mather, vice president of aerospace and defense industries at IFS.
Rob Mather, IFS’s vice president of aerospace and defense industries, speaks with Aviation Week Network about how the company is accelerating its artificial intelligence (AI) capabilities and how the industry might tackle potential roadblocks to the technology’s adoption.
IFS has been actively seeking out acquisitions of software providers with AI capabilities. Which areas of AI functionality or synergies are you looking for, and can we expect more of this action on IFS’s future road map?
It sounds a little trite, but IFS is all in on AI. We are going to continue to invest in AI both through internal development and through external acquisitions. For example, our recent acquisition of EmpowerMX not only brings with it many complementary pieces to our MRO solutions, but it also brings with it their EMX Vision capability that uses AI to help MRO organizations analyze their data to predict material availability and labor and skill requirements, plan more effectively and understand their business even better. It’s a very exciting AI capability to bring into our overall portfolio.
However, IFS isn’t interested in AI for the sake of AI. We are only interested in AI that adds genuine value for our customers in the industries we focus on, like aerospace and defense. We refer to this concept as Industrial AI, and currently, we’ve identified six general types of AI to focus on: content generation, recommendations, anomaly detection, optimization, event forecasting and prediction, and contextual knowledge.
We’ve also consulted with our company's industry experts and customers and validated more than 80 individual AI use cases that we are in the process of embedding into our system.
What types of AI functionality has IFS already incorporated into its MRO software portfolio?
IFS fundamentally takes a two-step approach to incorporating AI into our software portfolio. First, we embed AI capabilities in the platform and then we apply them to specific use cases to solve specific industry problems out of the box. As part of step one, AI capabilities become available for our sophisticated customers to make use of as tools that they can train or apply themselves. We’ve already embedded capabilities in all the focus areas I mentioned such as anomaly detection, event forecasting and prediction, contextual knowledge, and optimization. As part of step two we have applied our AI tools to solve particular challenges. These solutions, for example, include pre-trained AI models, and they follow generally two camps: generic features and industry-specific features.
We have many AI use cases available in the general functional areas available to our MRO customers. For example, our IFS.ai Co-pilot is available for software and process help within IFS Cloud. We have many predictive models for payment delay and deal close in our sales area. We have multiple AI-based solutions in our supply chain, like demand forecasting, that our MRO customers can use and get real value out of.
For MRO-specific solutions, we can make use of predictive maintenance modeling for assets as well as maintenance scheduling optimization. The list is extensive. We release new software every 6 months and with each release we add new AI use cases from the list.
What kinds of benefits do you think AI could provide for MRO?
When we talk to our aviation maintenance customers, they tell us that their number one challenge remains the skilled workforce shortage. According to a presentation late last year by Diana Santiago and Joe Pergola, with all current programs, we still aren’t expected to match the demand for technicians until 2040. Therefore, anything we can do to help technicians be more efficient, to be able to spend their time more efficiently and therefore accomplish more in the time they have, is a huge benefit.
Many of the biggest benefits with AI that we can achieve right away are around supporting technicians, helping them spend more time actually doing the maintenance instead of looking up information in manuals, waiting for parts, moving between locations and trying the wrong thing. AI isn’t coming for the technician’s job any time in the foreseeable future—we can’t replace the versatility of the human. Instead, it can be used to increase the quality of their experience when otherwise they are in danger of being severely overworked for the foreseeable future.
Do you think it will be difficult to get aviation regulators to approve AI technology for maintenance? What are potential roadblocks?
Just this past July, the FAA released its Roadmap for Artificial Intelligence Safety Assurance, following similar, earlier releases from the European Union Aviation Safety Agency and other regulators. In the roadmap, the FAA outlines the concept of AI as a tool. There are many ways to use AI today for MRO within existing regulatory frameworks. Generally, as long as AI is acting in a supporting capacity and all ultimate actions remain with the human, we’re okay. It changes, though, when we start to look at having AI make the decisions.
I used to be a big proponent of explainable AI for application in aviation. An oversimplification of explainable AI would be that because of the way you build the AI, you can always know exactly why the AI takes a particular path and comes to a particular conclusion. The reality, though, is that this practice is expensive and frankly limiting. I now believe explainable AI is key in certain areas but not others.
Instead, folks like Mark Roboff, the chair of the SAE G-34 Artificial Intelligence in Aviation committee, have brought my thinking around to the practice of AI testing. When we certify a human being, we don’t look inside their head to understand their reasoning. Instead, we train them according to specified guidelines and then we test them. If they pass their tests, they become certified. We need to implement a similar sort of testing and certification regime for AI in aviation in situations where the AI is being asked to make safety-critical decisions. If the AI passes the required tests, then it would be allowed to be used in the field—it gets certified. If it fails? Nope, sorry. Try again next time.
How do you think companies can get buy-in from maintenance technicians on the shop floor? How do companies convince them that AI is a helpful tool rather than a technology that might eventually replace them?
Bear in mind that almost all of us are already using many forms of AI today—sometimes knowingly, like when we ask ChatGPT or even just Siri or Alexa a question, or potentially unknowingly, like when our internet history is analyzed for targeted ads and offers; we use Google Maps or Waze to optimize our navigation route for the shortest time in traffic; anytime a streaming service like Spotify or Netflix makes you a suggestion on what to watch or listen to; or even if you use facial recognition to unlock your smartphone. AI has been playing in the background for years and if we continue to use it in a supporting capacity, just in more ways, it will make our interactions and experiences better.
But let’s look at the issue a little bit differently: as I mentioned before, we are facing an unprecedented shortage of technicians and all the estimates I’ve seen point to us only balancing out at the earliest in 2040. So, to keep planes flying, technicians are being worked hard because companies need to do more with the people they do have. What that means is that for the foreseeable future, all the AI tools are really going to do is help technicians to do more things more efficiently and actually make their lives easier—not take their jobs. No combination of AI and robotics is going to replace the flexibility of a human being any time soon, so there is no danger of technicians being replaced for a long, long time.
Now, how do we make them want to use it? The answer is twofold. First, it needs to genuinely help them with activities they find difficult or annoying today, such as looking up procedures in manuals, troubleshooting certain faults or doing manual data entry.
Second, it needs to be non-invasive. The user experience and the design of the interface need to be done well. It must work within the processes people are already doing to make those processes smoother. But it also can’t be annoying. It can’t be a “Microsoft Clippy” popping up when you don’t want it and giving you poor suggestions. It needs to add value, yes, but it also needs to feel like it’s adding value. If we do that well, then technicians will be clamoring for AI instead of resisting it.
Editor’s Note: You can read more about Mather’s perspectives on the most promising use cases for AI in MRO here.