At a glance

Drone logistics company Swoop Aero faced delays in understanding core operational data as it relied on technical staff to manually run complex SQL queries. DiUS developed a generative AI-powered text-to-SQL solution, enabling Swoop Aero’s teams to ask questions in plain English and get real-time insights through a simple chatbot interface. This streamlined data access and empowered cross-business collaboration, allows Swoop Aero’s business team to make faster, data-driven decisions without needing technical support.

The challenge

Breaking data bottlenecks to enable faster decisions

Swoop Aero’s drone logistics platform gathers extensive operational data on flights, equipment, customer service, and software development. However, accessing this data depended on technical staff manually running SQL queries to extract the necessary data. This dependency creates several obstacles to accessing and managing data efficiently:

• Bottlenecks in workflows, as non-technical users had to wait for technical teams to retrieve data.
• Slower decision-making, as data was not easily accessible when needed.
• Technical team overload, with staff constantly occupied with routine data requests instead of focusing on higher-priority projects.

Swoop Aero needed a solution that would allow all users—regardless of their technical expertise—to access data and gain insights on demand, without relying on technical staff.

What we did

Empowering teams with generative AI-driven data access

Swoop Aero partnered with DiUS to design and implement a generative AI-powered text-to-SQL solution, built using a combination of AWS services. 

“Working with DiUS, we developed a generative AI solution that integrates seamlessly with our existing AWS infrastructure. Their expertise ensured that we could quickly leverage powerful AI capabilities to enhance data accessibility and decision-making across the organisation.” Andrew Thomas, CPO, Swoop Aero

A generative AI-powered text-to-SQL solution enables non-technical teams to access critical operational data with simple, natural language questions. The need for SQL expertise is removed, as users can simply ask questions in plain language, like “How many flights were completed in the last month?” and receive data from the database. This reduces wait times for technical support and improves business agility. 

The core of the solution is a chatbot interface, which allows users to interact with the system using natural language. Amazon Bedrock provides the foundational generative AI capabilities and a serverless approach triggers the execution of SQL queries. The solution was built in two phases to enhance accessibility and functionality over time.

Phase one: Laying the groundwork for faster decisions

The first phase focused on creating a solid foundation for data retrieval by automating SQL query generation. To simplify SQL query generation, the team developed a system that automatically converts natural language inputs into SQL queries. Technical users who were familiar with the Amazon RDS Postgres database structure tested the solution by validating the generated SQL queries, improving accuracy and efficiency. Users now had a fast and reliable way of generating SQL queries that would give them the data they needed.

Phase two: Empowering non-technical users

Once the core system was in place, the focus shifted to making the solution accessible to non-technical users across the company and improving performance. In this phase, several key features were introduced:

• Self-correcting SQL: The solution could automatically detect and fix common errors in the SQL queries it generated, reducing the need for technical oversight.
• Context awareness: The solution was enhanced to remember the context of previous queries, allowing users to ask follow-up questions. For example, a user might ask, “What were the flight statistics for 2023?” followed by, “How does that compare to 2022?” The system would understand the follow-up without needing a complete restatement of the original question.
• Scalability and performance optimisation: As Swoop Aero’s data needs grew, the system was designed to scale efficiently. By optimising the way data was retrieved, the system was able to focus only on the most relevant information, speeding up data access and reducing processing costs, which allowed teams to work more efficiently.
• Making system tuning easier: The system enables non-developer technical team members to refine prompts and schema descriptions, improving chatbot performance. This allows the Swoop Aero team to get faster feedback on potential improvements without utilising developer time.

Through this two-phase approach, the solution evolved from an efficient tool for technical users into a powerful, user-friendly system that made data accessible to the entire organisation.

Results

Unlocking productivity and scalability with insights on demand

By enabling a self-service model for data retrieval, Swoop Aero has optimised operations, reduced the burden on technical teams, and increased the speed at which insights can be acted upon.

Now, the business team can easily answer ad-hoc data questions beyond what dashboards provide, reducing delays in reporting and enabling faster responses. For the Operations Manager, the solution delivers instant insights into performance metrics—such as identifying the pilot with the most flights or the most frequently flown locations—helping to improve operational efficiency. Meanwhile, the Production Manager can track part movements, pinpoint stop locations for specific components, and monitor team performance, allowing for better resource management and workflow optimisation.

“With DiUS’ expertise, we’ve built a dynamic system that leverages Amazon Bedrock and DynamoDB to deliver real-time data insights. The solution continuously adapts through user feedback, ensuring that our teams have instant access to the information they need.” Andrew Thomas, CPO, Swoop Aero

Currently, the generative AI-powered solution provides access to data in Swoop’s relational database, with the potential to integrate additional data sources for broader business applications. A key advantage is its ability to support role-based access control, allowing data access to be defined based on user roles. For instance, an Operations Manager could be restricted to retrieving country-specific data, streamlining access and improving the relevance of insights.