Stockland has built an AI assistant that connects about 700 property developers and project managers to procurement and purchasing data held in SAP.
Head of technology Sebastian Gray told the AWS Summit Sydney that while the data is important to decision-making, it is infrequently accessed and takes five or six minutes to served up each time.
“When people go to make decisions, often the info that they need to make those decisions is already present in the systems they have access to – it’s just it’s locked behind some sort of usability barrier, or complexity or technical awareness that they need around cost centres or purchase orders or security codes,” Gray said.
For example, when querying a purchase order, a user would have to “log into SAP, navigate to the purchase order management tile, put in a PO number/invoice/supplier number, execute the transaction and then finally they get to a result set,” Gray said.
“None of that sounds particularly difficult and, honestly it’s not that difficult if you know what you’re doing and you do it regularly.
“But when you go through that process from a standing start to actually doing it, it’s probably a five minute set of activities.
“When we look at querying project budgets, it’s actually a bit more complicated. There’s a range of codes people need to remember, project names, [and] financial details associated with it. This one probably takes them about six minutes to go through that process.
“When you add it all up, we’re talking about thousands of hours of effort from our people just trying to find the information they need to do their job.”
The company has now been an “intelligent AI assistant” that sits in Microsoft Teams to connect project managers and property developers with data in SAP.
The assistant, called Sage, took 18 weeks to build and launch into production after the business case for it was approved.
“[Sage is] an agent created and runs on [Amazon] Bedrock but it lives in Teams, [and] people access it through Teams. If you’ve already got an SAP login and Teams, you’ve already got access to Sage within the organisation,” Gray said.
“It allows people to just log in, start talking in natural language, and Sage then speaks ‘SAP’ [to SAP] and gets the actual data back for you.”
Hosting it inside of Teams was about “meeting people where they work”, Gray said, adding that people may be working in the field, at the project site, when they need the financial information.
It also meant the interface was relatively familiar to the target user cohort.
“This was not a new application, it didn’t require a whole set of new training, it was available to people on their phones where they needed it, inside Teams that they’re already very familiar with, in an interface that was just very obvious in terms of the way they should use it,” Gray said.
Gray said early usage figures were encouraging.
“In terms of outcomes, 700 users who generally don’t log into SAP have been using this naturally since we’ve gone live,” he said.
“We’ve had over 12,000 interactions with the tool in just a few short months, and we’ve seen a 95 percent helpful rating in terms of the way people are actually responding to using the tool.”
The company has built an AI agent to monitor interactions and alert the technical team if an information request problem is encountered.
“We get feedback the moment someone asks Sage to do something it currently can’t do, and that drives directly into our backlog of opportunities,” Gray said.
The company is now considering how it can expand the tool to other information retrieval use cases and domains.
Ry Crozier attended AWS Summit Sydney as a guest of AWS.
