Why The Cheesecake Shop CIO is calling a 34 year old business a startup as it rebuilds for AI

Retail is being redefined by intelligent systems, from edge analytics and AI-enabled networks to modular tech stacks and predictive infrastructure. 

But ADAPT’s research shows that legacy tech stacks remains a significant constraint on innovation and AI scaling across the sector. 

Many retailers continue to struggle with fragmented, ageing stacks that limit their ability to integrate customer, supply chain, and operational data, an essential foundation for personalisation and omnichannel optimisation. 

One organisation addressing this challenge is Australian retailer, The Cheesecake Shop. 

Established in 1991 and now operating 250 locations, the business has been held back by legacy systems and unstructured data. It is now building a foundation to modernise and make use of AI technologies. 

CIO Brad Dight tells ADAPT that the retailer’s platforms were entirely custom-built, with little deployed ‘out-of-the-box.’ Some tools, he says were implemented in ways that ultimately failed to deliver business value. 

“We’re taking this business and treating it as a 34-year-old startup pretty much, so trashing everything [tongue in cheek]. Not to say that predecessors [implemented] the wrong systems, it’s just that we’ve outgrown them and now we really need to enable the business”, he says.  

Brad explains that the organisation relied on a single database with no redundancy, creating a major point of failure. A single change on its intranet could have brought down the whole franchise network. 

To address this, the retailer is investing in new data warehouses and cleaning up its data structure before introducing AI capabilities. 

“There’s no point trying to overlay AI and…getting the wrong outputs. What we are not trying to do is create these really big tools and implement AI platforms to streamline the business. We’ve just got to get everything clean under the hood before we apply [AI].” 

Streamlining customer feedback and complaint processes

Brad says CEO, Scott Bush, is now challenging every department to identify practical AI use cases. One priority area is customer feedback and complaints management.  

“Every retail business does [receive] customer complaints; that’s part of the business whether you like it or not. We’ve partnered with HubSpot to look at that side of the business. 

 “One of the key things we need to focus on is speed to turn that around because [legislation dictates] that we’ve got to reply [to complaints] within a certain time. It’s about using AI not as a chatbot but being really intuitive to how it actually works.” 

For example, AI could automatically retrieve allergen information from the retailer’s website and send it directly to a customer, reducing manual effort and response times. 

“That streamlines the process and allows the complaints team to focus on the bigger issues”, he says. 

An agentic future

ADAPT’s data shows that agentic AI adoption in Australia is concentrated in operational automation, IT/security workflows, and data-driven decision-making. 

Most use cases remain in early-stage pilots or narrow production deployments rather than fully scaled autonomy. 

While AI tools are increasingly used for software development, true agentic coding (where autonomous agents identify, fix, test and deploy code changes), remains limited and mostly experimental. 

Brad says his tech team is ‘super lean’ and being able to identify software issues and automatically remediate them with agentic tools is the way forward. 

“Instead of engineers trawling through logs, the goal is self-solving systems”, he says. “We’ve got really smart people working on platforms. They can find what [the problem is], auto remediate it, test it and deploy it so they can get on with [higher value tasks].”