When a $30 Million Transformation Program Needs to Transform Mid-Flight

After investing $400 million in a manufacturing and distribution facility at Swanbank in southern Queensland and $30 million in a supporting digital and IT transformation, Suntory Oceania enjoys 38 per cent year-on-year growth, the number one position in the spirits market, and operational benefits out-delivering expectations.

But AI could have derailed it all.

When generative AI arrived halfway into the three-year program that united Suntory Beverage & Food and Suntory Global Spirits as one business, Suntory held its SAP scope and ran a second data program alongside it. Twelve months post go-live, the business runs conversational AI on its own data instead of waiting for a second transformation.

The physical build carried on unaffected, but the SAP platform designed and delivered by DXC Technology had to absorb a technology shift that did not exist when the scope was written. Phased against the commissioning of the plant, the go-live date was set in concrete.

“When we started our transformation three years ago, AI wasn’t really part of the conversation at all,” Matt Dixon, Digital & IT Director at Suntory Oceania, told an SAP briefing in Sydney. “So, we had to pivot during our build.”

Dixon separated the two problems. The transformation stayed locked to its minimum viable product while a parallel program moved legacy data into Databricks: nobody could yet name the AI use cases, but their demands on the data were already evident. That migration is now about 90 per cent complete. Early work rebuilt the reports the business already trusted; current work adds the granularity and semantic context that let everyone from the CEO to the shop floor trust the numbers, with Databricks Genie live over those datasets and Salesforce integration underway.

Readying the Workforce on the Same Logic

These tools pay off in experienced hands, Dixon says: knowing which part of an answer is worth keeping is what experience supplies. The rollout is staged accordingly, with a few hundred Copilot Pro licences this year, more next year, the web version for everyone, and global upskilling that includes leadership.

“You get one chance,” Dixon warns. “Staff used to seeing one number who ask AI a question and get a different number lose confidence and don’t come back.” Marketing can live with 80 to 85 per cent accuracy, where direction matters more than the decimal, but “anything transactional needs 99.99 per cent, otherwise it’s just not trusted.”

Paul Dearlove, SAP Practice Director at DXC Technology, sees the same problem across DXC’s clients: finance, supply chain, and marketing each building models with “slightly different datasets, slightly different models. Who’s got the source of truth?” He puts governance with the C-suite: “AI is a strategic tool, so it needs to sit with the C-suite, where strategy is decided.”

Foundations First, at DXC Too

Dearlove applies the same test inside DXC: AI tooling only pays off with someone who already knows the work. The 600-person SAP practice is being enabled one person at a time, starting with the customer delivery managers who already know the client, because without that guidance “it’s a wasted tool.”

Rather than tool recommendations, DXC gives clients a roadmap to a return: “the right foundations, like clean core, then SAP BTP, then AI foundations, then use cases with assistants before getting into agentic.” The 400+ SAP-approved Joule agentic use cases can then deliver quick wins in finance, supply chain, marketing, or HR.

Managed services faces its own disruption. “In the not too distant future, somebody or even an agent within a system will raise a ticket. That’ll get picked up by another agent. That’ll go into the repository, the corpus of information,” Dearlove said. “It will review that, it will write a fix, it will test the fix, and then it will put that fix into production.” The capability is not there yet, “but that will come to a world where that is zero touch.” DXC is recalibrating its go-to-market around AI-enhanced services, and Dixon is pressing his own suppliers at renewal to show how AI improves the service Suntory receives.

Where Suntory Becomes the Proof

DXC has held the Suntory account for more than fifteen years, knowledge that mattered for a date that could not move. Delivery ran in three phases aligned to plant commissioning, on an SAP S/4HANA core extended with EWM, PP-PI, Quality Management, and Plant Maintenance for the shift from discrete to process manufacturing. EWM integrates directly with the plant’s operational technology, including a 30,000-pallet automated warehouse on fifteen real-time interfaces, with the platform live and excise-ready for Australian bonded alcohol.

Year one delivered more than 32 million cases, a 36 per cent production lift, and 21 new products after the 1 July 2025 go-live. When New Zealand’s route to market launched on the same core, Suntory modelled 35 per cent adoption. It hit 70 per cent inside a fortnight.

Neither party treats the work as finished. Dixon is piloting agentic tools on trusted data “where we can scale them afterward. That’s when you will get a positive ROI.” It is Dearlove’s sequence for any client: foundations first, then use cases, then the return once it scales.

What This Means for Mastering SAP Insiders

Programs with fixed dates need new technology kept separate. Suntory ran the data work as a parallel program and protected the go-live rather than reopening the scope.

The data should move before the use cases exist. Nobody could name the AI use cases three years out, but their demands on the data were obvious, and the migration started on that basis alone.

Trust is won or lost on the first answer. Eighty-five per cent accuracy works for a marketing signal and fails for a transaction, so teams should set the bar by use case and build the semantic layer before people start asking questions.