Team Global Express has the logistics for AI use in place

Team Global Express has identified 73 use cases for AI across its operations, with 12 AI agents already in production.

Image credit: Team Global Express

The company, which spun out of Toll Group in 2021, has spent the past three years building data foundations and practices that now enable it to move forward with artificial intelligence.

Data, AI and integration director Michael Farrar told the AWS Summit Sydney that the company has built a centralised but modular data infrastructure, largely on AWS cloud services.

Previously, every line of business had its own database “on different data warehouses”, with varying data quality and no standardisation.

For tasks such as tracking a parcel, for example, that meant accessing multiple different systems and manually reconciling the transit movements.

“We knew that between [lines of business having their own infrastructure] and not having a single unified view, it wasn’t the sort of foundations we could build AI on, so we invested in centralising all of that data within AWS, moved to Redshift for our data warehouses and DynamoDB for our operational data stores,” Farrar said.

“From that, we’ve then driven out that single pane of glass. We can now see where our parcels are across any system.”

Data hygiene efforts are continuing, with the company progressively adding metadata and contextual data to stored records to make querying the data easier.

But the centralised data infrastructure, combined with a concerted effort around creating guardrails and security, has enabled the company to start adopting AI.

“We recognised AI could help enable our growth without increasing our cost in proportion to that growth,” Farrar said.

Out of 73 use cases identified for AI so far, Team Global Express has managed to get 12 “AI agents” into production. It also has a further five proof-of-concepts underway.

Farrar briefly described three of the production use cases.

The first uses Amazon Rekognition, an image analysis tool that processes proof-of-delivery images snapped by drivers when they make deliveries. The tool strips out any personally-identifiable information visible in the image, such as “the house number, names and addresses on parcels.”

A second use case, built on Amazon Bedrock AgentCore, aims to unlock operational intelligence for frontline staff that was previously inaccessible in various spreadsheets and databases.

“What that means from a benefit perspective is that employees now don’t need to go and create a BI dashboard that could take days or even weeks … if they’re trying to investigate a problem or an opportunity,” Farrar said.

“They can do that in seconds with AI. They can do an aggregation, a summary, a graph or … drive down into some detail.”

One of the most recent use cases to make it into production uses Amazon Connect to analyse the intent of callers to the company’s contact centre.

“Our next steps are then going to be to then use AI agents to start to answer some of those queries like, ‘Where is my parcel?’” Farrar said.

One thing helping Team Global Express to identify use cases is a team of 15 “AI champions” that sit across the business identifying opportunities and reinforcing data quality standards at a line-of-business level.

“They are SMEs [subject matter experts] for that business area. They know the business inside and out,” Farrar said.

“The key thing we did was partner with them. We went on ride-alongs with our drivers, we sat with our call centre agents, understanding what actually happens, not just what’s documented. 

“That was key to understanding what the use cases were - not just the easy ones the AI champions initially thought of because they knew AI could do it, but what are real problems and opportunities in that space.”

The company is using “hybrid” of out-of-the-box and custom AI products and services. 

It is allowing some “citizen development” of AI models, and has set some guardrails both to encourage safe development, and to ensure the code is production-ready at the end.

Farrar suggested the company may look at Kiro, a newer AWS service for enabling the development of software by non-coders using AI tools, as a potential “next step” to what it has currently put in place.

Ry Crozier attended AWS Summit Sydney as a guest of AWS.