How Domain-Specific AI is Transforming Aviation Operations

Inside Aerogility’s Approach: An Interview with Tom Godfrey on Domain-Specific AI

Generic AI solutions promise universal answers, but aviation’s complex operational realities demand something more sophisticated. In this Q&A, we explore how Aerogility’s domain-specific language approach is revolutionising how airlines, aviation businesses and any complex asset optimise their operations, uncover hidden inefficiencies, and drive meaningful change.

Before we dive into the questions, let’s introduce Thomas Godfrey. With a PhD from King’s College London in digital twin modelling, Tom brings a unique background spanning healthcare, finance, and academia. His earlier work with Domain-Specific Modelling Languages (DSMLs) was in healthcare, where he collaborated with NHS England hospitals to model and improve their response to COVID-19 in A&E. Just like aviation, this was a safety-critical environment where explainability, accuracy, and close collaboration with domain experts were essential. Today at Aerogility, Tom is passionate about bridging theory with practice; making AI more explainable, accessible, and impactful by working directly with customers to co-create flexible simulation solutions that solve complex challenges and deliver real-world results.

The Foundation: Understanding Domain-Specific AI

 Tom, let’s start with the basics. When we talk about “domain-specific language” in the context of Aerogility’s AI, what exactly does that mean, and how does it differ from generic AI approaches?

Typically, AI simulation tools rely on generic form-based interfaces for defining models. Users are required to manually translate their domain/business knowledge into an (often unfamiliar) platform-specific format to determine how the model should execute, rather than focusing on what they would like to model.

Collaborative Model Development with Aerogility DSML

Domain-Specific Modelling Languages (DSMLs) purposefully abstract away from technical complexity and instead reuse the language and format that business stakeholders are familiar with to construct simulation models.

In Aerogility, that means building models around high-level agents such as aircraft, components, maintenance centres, and operational tasks—rather than low-level technical entities and protocols. It makes the whole process much more natural and collaborative, because the language matches the way our clients already think about their operations.

You’ve mentioned that domain-specific language makes Aerogility “very accessible” for developing models directly with clients. Can you walk us through what this collaborative development process looks like in practice?  

Absolutely. It typically begins with a workshop, where we sit down with the client to thoroughly understand their business priorities and the challenges they aim to address.

From there, we work together to map their processes into an Aerogility model—for example, defining asset types, components, operations, and maintenance requirements. Once that’s in place, the model is executed as a detailed agent-based simulation, with each agent following sophisticated behaviours in the background.

The real advantage of Aerogility’s domain-specific language is that it keeps this complexity behind the scenes. Clients interact with the model in a way that feels natural to their business, while still benefiting from the depth and accuracy of the simulation.

In your experience, what are the fundamental limitations of generic AI when applied to aviation operations? What specific challenges does the aviation industry present that require this specialised approach?

As we all know, aviation is an incredibly complex and diverse domain, which makes it essential to define simulations that are both accurate and easy to understand. Safety and defence are critical areas, so it’s vital that simulation results can be explained clearly. With a domain-specific modelling language, we can directly show how inputs map to outcomes, making the process transparent. This is not possible with models defined in generic languages, where technical terminology and detail obfuscate the mapping from observed domain processes to appropriate model definitions.

The Discovery Process: Beyond Model Implementation

You’ve also noted an “often under-appreciated benefit” of Aerogility’s implementation process – the learning potential discovered during model development. Can you share a specific example of how this collaborative analysis has uncovered unexpected operational insights for a client? 

Defining a model requires careful inspection of domain processes. Doing so using a domain-specific modelling language allows users to take part in simulation development, opening opportunities to identify and investigate challenges that may otherwise not be exposed.

For instance, when working with a client on a maintenance program for a new aircraft specification, we discovered that many components would need inspections within a very short timeframe. This could have led to repeated grounding of the aircraft, something that wasn’t obvious until we mapped their maintenance plan into an Aerogility model using the DSML. This collaborative approach highlighted potential bottlenecks early, giving the client actionable insights before real-world issues arose.

How does the process of building domain-specific models force both your team and clients to examine their business processes more critically?

The impact of developing DSMLs within our own Aerogility team is an important one: in order to implement an effective DSML, we must establish a good understanding of our target domain.

Often, this will involve engaging with our clients to understand their business needs. Not only does this improve relations, but it also ensures that our technical teams are steeped in the aviation domain and can therefore more meaningfully guide clients in effectively using Aerogility to solve real-world issues.

When you’re working through a client’s operational challenges using Aerogility’s domain-specific approach, what’s the typical progression from identifying issues to testing potential solutions through simulation?

We start by spending time with the client to understand the full scope and depth of the challenges they face. We identify which elements of the domain are affected by the problem, as well as any potential ideas the client may want to test in addressing it.

Using the Aerogility DSML, we then help the client build a baseline simulation scenario that reflects their current real-world operations. This scenario can be validated by comparing real data to the simulation output, ensuring that the model’s predictions are accurate.

Next, we develop alternative scenarios that reflect potential solutions or interventions the client wishes to evaluate, measuring the impact of these changes on performance indicators such as availability and throughput. Importantly, the simulation can also reveal any unexpected consequences of these interventions, such as increased costs or knock-on delays.

Technical Deep Dive: Agent-Based Simulation and Language Design

From your PhD research and international expertise in agent-based simulation, how do you see domain-specific languages enhancing the power and accuracy of these simulations in aviation contexts?

I see the main impacts of DSMLs as the following:

How Domain-Specific Languages Enhance Aviation Simulations

• Improving the explainability of AI models. Many AI approaches remain a black box in which the methods by which outputs and decisions are reached are highly obfuscated. How can we know when we can trust these outputs, especially for safety-critical systems? The Aerogility DSML puts explainability and accessibility at the forefront, offering the tools for users to modify or implement their own models without needing to learn a whole new, unfamiliar language.

• Reducing the risk of misinterpretation and model inaccuracies. By exposing model details in a DSML, clients can directly analyse the details of a model definition and identify inaccuracies. This is not possible in generic AI modelling tools, where technical experts must manually translate domain processes to technical definitions.

• Improving client ownership of models. Clients can be actively involved in model development and can have full autonomy in developing new scenarios and experiments without relying on a third party to develop models for them.

Can you explain how Aerogility’s domain-specific language captures the nuances of aviation operations that would be lost or oversimplified in a generic AI system?

By exposing model definitions in familiar and domain-appropriate language, it is possible for users to engage more closely and therefore meaningfully in the simulation construction process. This closer collaboration offers increased potential for stakeholders to input more nuanced procedure details that overwise may be obfuscated in generic simulation platforms. The agent definition panels within Aerogility effectively then act as a ‘checklist’, exposing the set of parameters or details that a user may (optionally) want to include in a given model.

Real-World Impact: From Simulation to Results

You mentioned that clients can simulate different scenarios within Aerogility to investigate potential improvements. Can you walk us through how this scenario testing translates into actual operational changes and measurable results?

By running a set of scenarios in the simulation, clients can explore the potential impact of different approaches in a safe, low-cost, and controlled environment. This allows them to test both the benefits and possible drawbacks of proposed changes without disrupting real-world operations.

Once the simulation identifies the most effective solution or strategy, clients have solid evidence to guide real-world decisions. This not only reduces risk but also provides confidence that changes will improve key operational metrics such as efficiency, availability, and throughput. In many cases, scenario testing also uncovers insights or consequences that might not have been anticipated, giving clients a more complete understanding of their operational landscape before taking action.

As AI continues to evolve rapidly, how do you see the role of domain-specific approaches like Aerogility’s becoming more or less important in specialised industries like aviation?

AI is developing at an incredible pace, and the range of tools available today is very exciting. At the same time, it’s a period where caution is essential—we need to ensure these tools are used correctly and effectively. In industries like aviation, the cost of mistakes can be enormous. That’s why it’s so important to use tools that are trustworthy, explainable, and that clients can take ownership of. Domain-specific approaches, like Aerogility’s, give clients that level of confidence and control, which generic AI solutions often cannot.

For aviation businesses considering AI implementation, what questions should they be asking to determine whether they need domain-specific solutions versus generic AI tools?

They should ask whether the solution allows them to:

• Reduce the need to learn an entirely new, unfamiliar language to build models, cutting down on training and specialisation requirements.
• Take ownership of their simulation models, enabling independent scenario development while still having support available if needed.
• Understand and respond to unexpected results. Unlike black-box AI systems, DSMLs emphasise explainability, allowing clients to trace exactly how and why model decisions were made.

Based on your experience leading this work internationally, where do you see the biggest opportunities for domain-specific AI to transform aviation operations in the coming years?

The greatest opportunities lie in areas where complexity and safety intersect—like maintenance planning, scheduling, and resource allocation. By combining domain expertise with powerful, explainable simulations, aviation businesses can optimise operations, reduce costs, and improve reliability in ways that generic AI simply can’t achieve. Looking ahead, I’m most excited about helping clients use these tools to uncover hidden inefficiencies, test innovative strategies safely, and make more informed, evidence-based decisions.

Call-to-Action

Ultimately, domain-specific AI isn’t just about building models; it’s about empowering teams to understand their operations more deeply, make smarter decisions, and confidently navigate the challenges of highly complex, safety-critical environments like aviation.

For readers considering AI in their operations, you might ask yourself:

• Could a domain-specific approach help my team uncover insights that are hidden in spreadsheets or manual workflows?
• Are our current tools obscuring how decisions are made or limiting our ability to test new scenarios?
• What hidden inefficiencies or risks might we uncover if we could safely simulate different scenarios before implementing changes?

These are the kinds of questions that can help your organisation evaluate whether a specialised AI approach like Aerogility’s could deliver real impact.

Check out other posts where you can meet the rest of the team or email us for more information on this story.