FlyBuys reveals analysts' virtual assistant "Nex"
Key points
- Flybuys has launched Nex, a virtual assistant built on Snowflake’s CoWork, to help marketing analysts make commercial recommendations.
- The data “pantry” built on Snowflake in AWS Sydney initially eased access to millions of daily transactions and signals, but this shifted the bottleneck to investigation, prompting the need for AI analytics.
- Linking Nex to metadata and business context was the breakthrough that fixed early plausible-but-wrong answers, though Flybuys is cautious about rolling it out beyond trained analysts.
FlyBuys has developed a new virtual assistant for its marketing analysts that is allowing them to use their time more effectively when making commercial recommendations.
The launch of the new assistant, called “Nex”, follows a program of work that was designed to make millions of transactions and signals generated by the loyalty program operator’s 10 million members daily more useable for analysts.
Speaking at a Snowflake customer event in Sydney FlyBuy’s head of data and insights Jane McCarthy said that work to establish the data foundation, which she described as a data “pantry”, was initially to allow analysts to access data in a timely way.
Flybuys deployed Snowflake in AWS Sydney and used its native data engineering, data lake, and AI to deliver the capability.
“Before the pantry, analysts were often constrained by access to information. So much effort went into finding the data, understanding the definitions, bringing information together. The analyst’s time was being constrained by lower value work,” she said.
However, the move changed analysts’ behaviour, snowballing into a need for AI analytics to increase their ability to investigate the data.
FlyBuys found that its analysts were asking increasingly complex and consequential questions.
“Improving access to the data didn’t reduce the need for analytics, it actually increased it,” she said.
“The pantry solved a lot of our traditional challenges, helped with finding data, understanding definitions, reporting consistent metrics, but it also unlocked more complex work. Analysts were able to connect more data points, test more hypotheses, explore more possibilities.
“The bottleneck shifted. It was no longer about finding information. The new bottleneck was investigation and this is where AI became interesting to us.”
FlyBuys developed Nex, a virtual assistant based on Snowflake’s agentic AI offering, CoWork. The idea behind Nex was to give analysts more confidence in their judgements rather than replace them.
McCarthy said that FlyBuys’ initial attempt to connect its data and AI “wasn’t great”. Nex was producing answers that were deceptively plausible but on closer inspection should have been received with a lower level of confidence.
“The fact that they seemed plausible is actually more dangerous than if they were obviously wrong, because people trust confidence,” she said.
The problem, she said, was that Nex didn’t understand the business context.
“The real breakthrough came when we linked Nex to business context; to our metadata. That really changed the conversation. Nex was no longer just retrieving information, it understood the business around that information [and] that fundamentally changed the quality of the interactions and the accuracy of the output.
“That was the holy smokes moment where it was really clear how powerful this is,” McCarthy said.
McCarthy said that FlyBuys considered making Nex available across the business beyond analysts but ran into some risk concerns.
For now, she said, the company was “slowing down and acting very cautiously”.
“With our analyst users, they’re trained to question, they’re trained to validate. They can look at the query Nex produces and sense check it, validate that it’s actually going to reach an answer that lines up to the question that we’re applying it to.
“Business users won’t necessarily have that capability,” she said.
About Author
