Session

A Different Way to Compete: How Great Southern Bank Went All In on Data and AI

Overview

Experience	In Person
Track	Analytics & BI
Industry	Financial Services
Technologies	Genie
Skill Level	Beginner
DOWNLOAD SESSION SLIDES

Great Southern Bank (GSB) has successfully migrated from SAS to Databricks - unlocking higher model accuracy, faster insights, and the ability to democratise data at scale using Genie and AI/BI. Importantly, GSB achieved this with a much smaller team than the major banks, demonstrating practical expertise and delivery experience that other organisations can learn from. To date, GSB has migrated and integrated 85% of its enterprise data onto the Databricks platform and is on track to shut down its legacy warehouse by 2026. Databricks AI capabilities are already delivering material value across traditional ML and gen AI use cases - e.g. a home-loan churn model was built by a single employee and matched the accuracy of a 3rd party model. On top of the enterprise data foundation, Genie Spaces now enable business users to self-serve insights faster and with improved accuracy. GSB is also accelerating value creation in areas such as complaints analysis and sentiment using GenAI.

Session Speakers

James McNiff

/Senior Solutions Architect
Databricks

Matt Cammack

/Head of Customer Technology Data and AI
Great Southern Bank

Full Summary

How Great Southern Bank turned small scale into a data and AI advantage

Great Southern Bank, a customer‑owned Australian bank with roughly 1,000 employees, chose to compete on speed and intelligence rather than size. Leadership abandoned incremental fixes and rebuilt the data stack on a single governed platform so every new use case compounded value instead of fragmenting it. The story offers a pragmatic blueprint for midsize institutions weighing evolution versus a clean start on a Databricks Platform. For financial services firms navigating similar decisions, the Databricks Platform for Financial Services outlines how governed, scalable foundations unlock lasting competitive advantage. After laying those foundations, the conversation turns to a practical question many enterprises face: with more AI ideas than capacity, how do you choose what to ship first so value shows up fast and visibly?

Why a small bank started over on data

By 2021 the bank's data estate had grown organically into multiple independent warehouses, fragmented ownership, and spreadsheet‑heavy workflows. Conflicting metrics and lengthy fact‑checking slowed decisions. Crossing AUD 20 billion in assets tightened regulatory expectations, raising the bar for clean, trusted, traceable data.

Leaders opted for a ground‑up rebuild on the Databricks Platform, standardizing governance from day one with Unity Catalog for lineage, fine‑grained access controls, and traceability. The aim was a strategic asset that could scale governance, reduce complexity, and make intelligence routine across the bank.

One governed platform, compounding outcomes

Three legacy platforms were consolidated into one enterprise environment. Pipelines, testing, and monitoring were automated. As business teams transitioned, manual spreadsheet work was simplified and historical control issues were resolved. Every function, from product and customer to finance, credit, risk, capital and regulatory reporting, now runs on the same governed foundation.

Measured impact arrived in waves:

Regulatory returns that once took days across multiple people became fully automated and completed in hours by 2023.
Embedded data quality and monitoring reduced fraud investigation effort by over 15 percent by 2024, shifting time from collation to investigation.
Higher data confidence unlocked additional lending capacity equal to about 2.6 percent of the loan book by 2025 through improved capital allocation.
The program achieved full payback by 2026.

AI value accelerated on top of the foundation. A single analyst rebuilt a third‑party home‑loan retention model in weeks on the Databricks Platform, matching and improving performance while returning the IP to the bank. Teams are now moving into scenario planning, forecasting and stress testing, and preparing first agent deployments to automate quality assurance. Perhaps the biggest shift is how people work: non‑coders use Genie and Genie Code to query live, governed data in natural language, bringing fresh analysis into decision forums instead of relying on stale offline reports.

From foundations to AI: picking first bets with discipline

Many enterprises experiment with AI, yet far fewer productionize. The gap often reflects prioritization, not capability. A simple filter helps focus effort where it moves the business: map candidate use cases by two factors, value and readiness. Prioritize the high‑value, high‑readiness quadrant first, invest to unlock items that are high value but not yet ready, harvest quick wins judiciously, and park the rest.

Self‑service analytics frequently lands in the high‑value, high‑readiness corner when data is already governed on the platform. Little net-new build is required, and benefits reach every function, creating visible momentum while reinforcing the same governance backbone.

Turning a retention fire drill into a one‑minute workflow

Mortgage retention illustrates the payoff. Identifying high‑risk customers with fixed rates expiring in 30 days typically spans requests across core banking, ML scoring, and warehouse teams, taking days while competitors pitch. Using Genie against Unity Catalog governed data, an analyst without SQL can ask the same question in plain language, then follow up by ranking the top customers by value and risk, with results in under a minute and permissions enforced automatically. Add model explainability and the retention team can tailor outreach to why each customer is likely to churn.

Compete on decision speed, not organizational size

Advantage now accrues to institutions that turn trusted data into confident decisions fastest. Treat governance as a strategic asset, standardize early to avoid fragmentation, and invest in how people work, not only in tools. A single, governed platform creates the conditions for AI to scale, and a disciplined prioritization filter ensures the first wins arrive quickly and keep compounding.

FAQ

Fragmented warehouses, manual processes, and inconsistent metrics made incremental fixes unlikely to deliver clean, trusted data at the needed scale, especially after crossing the AUD 20 billion asset threshold. A clean start enabled unified governance, automation, and reuse so value could compound.

Results arrived in phases. Automated regulatory returns landed by 2023, data quality monitoring cut fraud effort by 2024, improved capital allocation unlocked about 2.6 percent of lending capacity by 2025, and the program achieved full payback by 2026.

A governed data foundation, automated pipelines, and standardized tooling removed data plumbing and access hurdles. The analyst focused on modeling, matched and improved vendor performance, and returned the model's IP to the bank.

Score each candidate by potential business value and current readiness to deliver with in‑house skills. Start where both are high. Self‑service analytics often fits because governed data is already available and benefits are broad.

No. Queries run with the requester's permissions through Unity Catalog. Results include only data that the user is authorized to see, even when questions are asked in plain language.