Artificial Intelligence (AI) is having a positive impact on almost every industry. It improves decision making processes, creating fast and consistent operations management. In the specific field of Aerospace, our conviction is that to be fully efficient, AI must be developed with dedicated characterics. Sopra Steria invests on these features for sustainable and large scale transformation by AI for aerospace companies.
Hybrid Artificial Intelligence, a crucial decision support technology
To support decision, using hybrid Artificial Intelligence, which feeds on both data (e.g. images, historical transactional data, etc.) and the business knowledge of Supply Chain experts and operational staff is essential. This approach ensures to end-users a substantial analytical traceability of the AI’s decision-making and reasoning process, reducing any black box phenomenon.
This is particularly valuable since explicability of Artificial Intelligence systems is at the heart of manufacturers concerns, especially concerning critical systems. Hybrid AI, which is partly based on human expertise and modelling, is thus one of the ways to handle explainable AI. This is in line with Sopra Steria investments related to major institutional programs such as GAIA-X, ANITI and Confiance.ai, which aim to industrialize a trusted and sovereign Artificial Intelligence.
Hybrid AI can be also used to create relevant algorithms in "small data" environments, which is a situation often faced by aeronautics manufacturers. The industry also often needs to provide decision support in situations where historical data is non-existent.
Platformisation, a major lever to create value for Aerospace manufacturers
Decision support systems based on hybrid AI need to be industrialised on a large scale to create maximum value for aeronautics manufacturers. To achieve this, the notion of platformisation is a key. By integrating AI services onto a platform, deployment is faster and more sustainable over a larger business scope, while ensuring the cybersecurity of the target system.
There is significant importance on using platformisation as a tool to create awareness and extend the reach of AI systems. Platformisation is also a powerful lever to ensure the urbanisation of all AI systems with one another, especially thanks to the containerisation and APIsation of services. This approach brings together the coherence of data and knowledge, as well as the ability to create feedback loops incorporating the performance of decisions. Systems, too, are easier to monitor and supervise in production to control their relevance and performance regarding operational objectives.
Finally, the use of AI platforms is a way to manage the energy and resources consumption of the implemented decision support systems, allowing improvements with regards to the green efficiency performance of IT systems on the one hand (“green for IT”), and the carbon footprint of the Supply Chain Management on the other hand ("AI for green").
An iterative approach to implement AI within IT legacy landscapes
AI decision support systems can be grouped by families of algorithms, allowing modeling, data and knowledge management, highlighting or creating value through huge amounts of information that human intelligence simply can’t absorb.
These algorithms, first, take the form of Descriptive AI, answering the question, "What is my or my organisations’ current situation?” in order to give insights to the decision makers concerning tangible elements of the Supply Chain state and environment. The technologies considered will be close to business intelligence or advanced analytics algorithms.
These systems can then take the form of predictive algorithms or "predictive AI", this time making it possible to understand the evolution of the behaviour of the modeled system and thus to answer the question "what will happen?". Decision-makers can then understand the trajectory of the system they have to monitor and be able anticipate its performance. This will involve technologies such as machine learning, deep learning or Bayesian inference.
Prescriptive decision support systems, or Prescriptive AI, also have a role to play. This AI directly proposes recommendations to decision makers so they can to optimise problem resolution. This can be an explicit behaviour to adopt in order to respond to a specific problem or a choice to be made at the time. In other words, these decision-support systems answer the question "Given the current and predicted situation of my Supply Chain, what should I do to optimise its performance and avoid a snowball effect?”. From a technology perspective, the approach considered will use optimisation and operational research, reinforcement learning, or simulation.
If the decision recommended by the system is ethical, explainable and acceptable to humans, it is possible for the decision maker to let the system act and execute the recommended decision on its own. In this context, even if the technologies used are unchanged compared to prescriptive decision support systems, the level of AI considered is known as "Autonomous AI".
What are the next steps?
The outcome of implementing AI can be transformative, productive and – most of all – a seamless transition from labour-intensive tasks when done correctly. AI offers for the Aerospace Supply Chain exciting prospects that can be harnessed with the right guidance, approach and trust.