Building Ask Aida
Lessons from bringing conversational AI to CACI’s data
Discover how CACI built a powerful, secure, customer-facing AI interface for trusted data analysis – how we did it and how we support our customers.
Launched in June 2026, Ask Aida is CACI’s conversational AI interface for our consumer and market products and services. The challenge was simple: create a familiar, chat-style AI interface that can be used by anyone to surface CACI’s data.
We decided to build Ask Aida from scratch. It needed to meet the customer demand of making CACI’s data easier to access against the backdrop of a rapidly evolving AI landscape. It also needed to broaden access to data and insights to anyone in an organisation, irrespective of their data literacy.
Where Ask Aida differs from regular AI tools is that it offers access to CACI’s data and models, as well as enabling customers to layer in their own data when performing analysis. Type in a question, receive an answer. It sounds simple and familiar, doesn’t it?
Why CACI built Ask Aida
Our primary goal in building Ask Aida was to enhance and democratise access to CACI’s data, models and services. Low-friction access was created by providing an easy-to-use conversational interface. The aim was to make analysing data and drawing out insights easy for everyone, thereby removing bottlenecks such as needing to book in analysts’ time or updates to dashboards.
At the same time, our customers were requesting help and guidance in implementing AI in their businesses. “We thought, what better way to help our customers than having been through this whole transformation problem ourselves first?” explains Stewart Moody, head of product at CACI. “Our customers were having problems with this. If we could solve this ourselves quickly, and make AI work for us in practice, there is value to our customers in being able to share our learning.”
Generative AI is changing how people interact with data, but the foundations are familiar to us at CACI. For decades, we’ve combined rich data, advanced algorithms and technology to help organisations make better decisions. AI now gives us the opportunity to make that intelligence considerably easier to access and act upon. Our team had extensive AI experience, fitting it into workstreams across the business. This led to the realisation of the revolutionary potential of AI at an early stage.
Building Ask Aida: how we did it
“We were creating AI tools within our existing products,” says Stewart. “But the experience was still skewed by the user journeys of those products that were built before GenAI existed – heavy UI, then a simple conversational experience.
“When we were considering what the market was demanding, what users really need, and all the tools available, we decided to act like a start-up and simply build, from the ground up, what was right for the use cases given the tools and technologies we now have.
“Our team was faced with several options in achieving this. Starting out in April 2025, the AI landscape was evolving rapidly. There were myriad technology options and no set way of building and realising a project like Ask Aida.
“We had to make educated guesses on some of the tools we used. A good example is the team’s use of the model context protocol (MCP). MCP is now the de facto way of connecting AI agents and data sources. We experimented with it at the time, seeing that it’s a great tool that was gaining traction. It was a successful experiment. We played with a few other technology choices as well, some turning out better than others.”
Ask Aida: the project that never sleeps
Ask Aida is an ongoing project. It’s not a tool that we’ve built, then been able to sit back and admire our work. The AI landscape is still evolving, so the tech behind Ask Aida requires constant review. Where necessary, the team will replace them with new components that come to market and enhance Ask Aida. It’s a story of constant evolution and improvement.
”We quickly realised we could spend three years working out the perfect tool, only for it to be out of date by the time of release,” adds Stewart. “So, this start-up mentality has worked well. We productionised Ask Aida for internal users seven months after writing the first line of code, often using agents to write the code for us. We’ve then been able to adapt to new challenges and embrace new technologies rapidly as the project advanced. The same mindset exists within the team now.”
Testing for responsible AI use
One of the major challenges with any AI tool is the potential for hallucinations and inappropriate answers. To reduce these risks, our dedicated team of data and AI ethics experts, led by Sue MacLure, rigorously stress-tested Ask Aida’s responses for errors, bias and potentially offensive content.
From this, we implemented denied topics, which is critical to ethical use of AI. “The principle sounded great,” says Stewart, “but once we turned all this stuff on, Ask Aida became the biggest ‘computer says no’ system in the world. It was avoiding responding to legitimate questions. That took some fixing, but it was a good learning curve.”
Why CACI built Ask Aida
With the intention to build an AI tool with a familiar feel, what has set Ask Aida above and beyond the established names in the AI space? Why wouldn’t you just use a tool like Copilot instead?
Our team ran several side-by-side tests with Copilot during the development of Ask Aida. This was done to ensure that Ask Aida was hitting its targets and that it was offering value beyond publicly available data that other AI products typically have access to.
The team concluded that Ask Aida, with its unique access to CACI’s data, when asked to provide insight on consumer and market data, was providing more comprehensive responses than Copilot in 90% of instances. This ensured that Ask Aida was successfully surfacing CACI’s data in a meaningful, conversational and familiar way for users.
Trust and security first
CACI is Cyber Essentials+ certified, but we wanted to go beyond that standard in ensuring that Ask Aida is as secure as possible. “We worked with our American parent company’s red team and challenged them to behave like a hacker and see if they could get in and cause mischief,” explains Stewart.
“This proved to be a huge success, unearthing improvements we could make and implement to strengthen the security of Ask Aida and absolutely minimise the risk of external threats. We’re now able to take the learnings from that and apply them across the rest of our product portfolio, running security checks with our US parent to ensure that they are as secure as possible.”
AI foundations to support our customers
Trust sits at the heart of Ask Aida. Without trust, users lose faith in the responses and when that happens, they walk away from the product. From conception to delivery, we’ve thought about and designed Ask Aida as a trusted source of valuable insight to our customers. In providing a very simple output via a text interface, a lot of complex design and implementation has taken place in the background.
We’re proud of what Ask Aida has achieved. We’ve been able to deliver a tool which facilitates data analysis and democratises access to it. It looks simple, but the process behind it has been anything but.
From deciding on which technology solutions to use, ensuring appropriate responses and getting the security right, the Ask Aida project has been a huge success for CACI. It lays the foundations for future AI products and the ongoing support of customers in their own AI projects.
We’re excited to put our AI learnings and foundations into practice for our customers, working across the many industries and sectors they represent. Want to learn more about Ask Aida? Visit the software page.
