ANZ surfaces its internal software delivery aid, Ensayo AI

ANZ Banking Group has shed light on Ensayo AI, a generative AI tool it developed with HCL and AWS to make parts of the software development lifecycle, from requirements building to testing, more efficient.

Rico Zhang at AWS Summit Sydney.

Platform engineering and SRE capability area lead Rico Zhang told the AWS Summit Sydney that a minimum viable product (MVP) of Ensayo AI is now live in production at the bank.

In May last year, the bank first disclosed an exploration of GenAI to improve the efficiency, reliability and performance of its code, but there was little detail at the time on what form this might take.

Ensayo AI - which runs in Amazon Virtual Private Cloud (VPC) - has been trained on “thousands” of system specifications, product documents, regulatory requirements and previous build requirements.

The end result is a tool that can help teams at various stages of the software development lifecycle (SDLC) - “helping business analysts write better and more accurate requirements at the beginning; then translating those business requirements into technical build specifications for the engineer and answering their questions along the way”; and finally, writing test automation scripts that have led to sizable reductions in test times, Zhang said.

Zhang said Ensayo AI was needed in part to help technically capable engineers get to grips with “the complex world of banking”. 

“Suddenly you’re facing terms like credit hierarchies, risk profiles and maybe hundreds of acronyms that you haven’t heard before in the business requirements, and you start realising that you really need to understand those concepts before you can effectively design a solution and code it,” he said.

Zhang said that it could take two-to-three months of concerted effort to learn the language of the bank’s requirements so it can be reflected in the code. Still, misunderstandings of a requirement are possible, requiring additional work and lengthening the timeframe for delivery.

In true augmentative fashion, Zhang introduced Ensayo AI as “a new team member” - taking the form of a “GenAI solution that is equipped with ANZ banking knowledge as well as the engineering expertise” to improve software delivery.

Test automation

Among Ensayo AI's biggest benefits are in the area of testing, which Zhang said could take up “more than 50 percent of a project’s time.” 

“Other than unit tests, we have API tests, integration tests, functional tests, and multiple rounds of regression tests,” he said.

“Those are very important measures for us because as a bank we’re dealing with real people’s money, and we have very little room for mistakes. We have to make sure that the whole system and also the integrated system continue working as expected after your code change.”

Zhang said the bank had contemplated automated testing before, given it “could really help us”, but he said it “often got pushed aside because of deadlines, other priorities [and] urgent work.”

Aside from understanding complex banking requirements, Ensayo AI is built to “enable test automation with minimal cost”.

The results so far are extremely positive: “We were surprised to find out our API testing time has been dramatically cut by 72 percent and our integration testing time has been reduced by 56 percent, which greatly accelerates our project delivery,” Zhang said.

Ensayo AI's build

Zhang said that working with long-time partners HCL and AWS had been “pivotal” to realising the vision for Ensayo AI.

He said that the bank’s alignment with AWS as a cloud partner, and the regulatory acceptance of those arrangements, made the cloud service provider an obvious choice for GenAI tooling.

“Selecting AWS as our GenAI provider for Ensayo AI wasn’t a decision that we made lightly. You can imagine we tried a few different platforms, but AWS really stands out in a few very important measurements,” Zhang said.

“First of all, AWS is one of our strategic cloud providers, so based on the established governance framework, and a strong technical foundation, we have access to all of the necessary GenAI services and infrastructure within a couple of minutes.

“Second is for compliance - a big part of our job in the bank is to talk to the regulators. Even though Ensayo AI is not processing production customer or banking data, it’s still very important for us to make sure our regulators are happy with what we do with GenAI and with cloud. 

“So, the strong security foundation, the mature documentation, and the easy access to their professional services team really help us with those conversations.”

Zhang said that his team initially shared the concept for Ensayo AI with HCL. From there, things moved quickly: within two-to-three weeks, the three organisations - HCL, AWS and ANZ - were “sitting at the same meeting and [drawing up] the attack plan going forward.”

He added that the road to a production instance of Ensayo AI had been a multi-stakeholder journey for the bank’s internal teams as well, and many had direct involvement.

“GenAI is new to everyone, especially for highly regulated industries like banking, so to land a GenAI solution in a production capacity was really a journey inside of ANZ,” Zhang said.

“We have daily involvement with our architecture team, risk team, legal, security [and] data governance [functions]. Together, we designed and built this solution from day one.”

Zhang said the internal path had been smoothed with the support and sponsorship of the bank’s CTO - Tim Hogarth - and its CIOs.

“With this support, we achieved cut-through in a very short period of time,” he said.

Reference architecture

Zhang noted that since launching Ensayo AI, a pipeline of ideas and additions to the tool has emerged.

He also said that Ensayo AI is being held up within the bank of a case study for how to get a GenAI tool into production.

“If you look at GenAI as a whole topic within the bank, it can be a very complex problem. [Having a] use case with go-live, we provide a reference point for different use cases to look at how you can introduce AI in a controlled and secure manner,” Zhang said.

“[We’ve] become a reference architecture for a lot of use cases that [the bank] can build upon.”

Ry Crozier attended AWS Summit Sydney as a guest of AWS.