Databricks vice president and country manager for Australia and New Zealand (ANZ) Adam Beavis says the market will reach a tipping point in 2026, with artificial intelligence (AI) projects moving from the exclusive domain of engineers to a core element of the business environment.
In an interview with Computer Weekly on the sidelines of the Sydney leg of his company’s Data+AI World Tour, Mr Bevis said organizations needed to move quickly to replicate the success seen at companies like Suncorp. However, he cautioned, “Everything has to start with good data.”
Once that foundation is in place, you can quickly implement agent AI. Databricks aims to handle the “undifferentiated heavy lifting” so customers can focus on execution, Beavis said. He noted that business units are increasingly looking for ways to extract value from their holdings by visualizing value using tools such as Genie (the company’s intelligent assistant) and dashboards.
Beavis noted that there are “a lot of self-service” features within the platform, which business leaders are specifically looking for.
However, training is still required. The company hosts sessions called “Hackathons for Businessmen,” which bring together staff from various business groups to demonstrate the ease of use of the technology and bring it back to their daily work.
The company is currently reporting 70% year-over-year growth. Beavis expects future expansion to be driven by a shift away from legacy systems and a growing desire to utilize time-series data in the energy and utilities sector.
While energy has long been an important area for Databricks, the company is looking to grow financial services, AI-native startups, and organizations looking to combine operational technology (OT) data with broader business data through partners such as SAP.
Solving the data silo problem
Craig Wiley, senior director of product management at Databricks, suggested that while all industries stand to benefit from AI and “the way we work is changing,” many companies are struggling to take advantage of this technology.
A common roadblock is having data locked into multiple (often proprietary) systems with fragmented security and governance frameworks. This prevents employees from finding the data they need and prevents automation.
Databricks addresses this problem by using open formats to consolidate data into low-cost storage and applying an integrated, fine-grained governance approach on top of that. The platform also supports composable agents, simplifying the deployment of pre-built agents, custom organization agents, and traditional models used for tasks such as churn prediction.
Several Databricks customers shared their experiences at the conference.
Telstra: From crimes against data to AI first
Telstra has set a goal to become an AI-first organization. But Dale Stevens, the company’s data and AI executive, acknowledged that given the carrier’s long history, technical debt and past “crimes against data,” it needs to get its foundations right first.
Advancements include separating compute and storage to reduce costs and improve scalability, establishing data lineage, and enabling data sharing across multiple platforms and clouds.
Mr Stevens highlighted Telstra’s commitment to responsible AI. The company was an early adopter of the Commonwealth Government’s AI Ethics Principles and was the first Australian company to join UNESCO’s Business Council to promote the implementation of ethical AI.
“All of Telstra’s AI models go through oversight boards,” she said, with goals focused on protecting privacy and preventing AI from slipping into “creepy conditions”.
Culturally, Stephens noted that there is a tension between early adoption of technology and maintaining customer and employee trust. Our priority is to maintain that trust through transparency, data quality, and governance. To support this, Telstra has established a Data and AI Academy and 20,000 employees have already completed at least one course.
Fonterra: Data-driven dairy
Helius Guimaraes, chief data and AI officer at New Zealand dairy cooperative Fonterra, explained how the organization is using Databricks to support transformation.
Fonterra took a data product approach because much of its historical data is stored in legacy systems. These reusable products, combined with AI-powered self-service capabilities, deliver greater efficiency than traditional reporting. Fonterra currently has more than 6,000 employees using generative AI capabilities within a variety of applications.
Alinta Energy: Retirement of vulnerable models
Alinta Energy’s spot trading team previously relied on weak machine learning (ML) models developed more than a decade ago in the Matlab programming language and an analytical environment on a desktop PC.
Andrew Davis, delivery manager for data and AI platforms, explained that the old model “took the business days to train” and was only updated annually. By migrating to Databricks, ingestion and processing tasks that used to take hours can now be completed in minutes.
Andrew Gorkic, principal AI consultant at Fujitsu Australia (which provides services to Alinta), said Databricks also provides ML operational support. This automates “challenger vs. champion” testing and alerts you when the model’s output deviates from reality.
This change has reduced the stress on our trading teams, allowing them to focus on market trends. The new model is faster and more accurate because it can be retrained every 1-2 months depending on weather and market conditions.
Prospa: Mapping fraud using graphs
Prospa, a small and medium-sized enterprise (SME) lender, uses Databricks to approve microloans in under an hour. However, mapping the complex relationships between businesses and individuals required going beyond traditional databases.
Lead AI Scientist Jin Foo explained that by adopting Neo4j’s graph database, Prospa was able to better visualize risks. For example, if 10 companies with the same owners apply for loans at the same time, or if a fraud group uses dozens of shell companies, a graph database will reveal these connections.
Prospa plans to implement search-enhanced generation by connecting graph databases to large-scale language models, allowing operations teams to query data using natural language.
