The cloud giant is embedding cross-functional teams of engineers, scientists and strategists with customers, with Commonwealth Bank of Australia among those already seeing results
The move, unveiled at the AWS Summit in Washington DC on 30 June, is the latest phase of a strategy that began in 2017 with the creation of the Amazon Machine Learning (ML) Solutions Lab, according to Taimur Rashid, managing director of the AWS Generative AI Innovation Centre.
At the time, there was growing interest in applying machine learning to areas such as product recommendation and fraud detection, and the company had launched SageMaker, its cloud-based machine learning platform.
With an emerging technology and a relevant new product, AWS decided the best way to bridge the gap between experimentation and business value was to set up the lab and do what is now known as FDE.
The lab embedded scientists and data scientists with customers, where they could learn the business, understand the use cases, create prototypes and build working systems.
“What we saw over the course of the three years that followed was the adoption of machine learning by our customers,” he said. “Companies like Airbnb, Netflix and Sony were using machine learning to power various parts of their business, and AWS’s investment in expertise and forward deploying it for customers helped bridge that adoption gap.”
When generative AI took off in 2023, “we rebranded, we called it the Generative AI Innovation Centre, and put $100m worth of investment into the initial formation of the team” to support the deployment of strategists as well as scientists, because generative AI affects so many aspects of business.
“That really helped customers such as the NFL, the PGA Tour, manufacturing companies like Jabil and automotive companies like BMW, as it shortened the time for them to realise value with AI.”
The centre helped build a lot of prototypes, and customers then asked for help putting them into production, as well as for skills transfer.
“Around 2025, we made an additional investment into the innovation centre with another $120m to build on agentic AI, and that resulted in a number of customers such as Cox Automotive [the company behind Kelley Blue Book and other brands], State Street [a US-based banking and financial services company], Itaú [Latin America's largest bank] and Visa all leveraging agents in production across many of their internal and external workflows.”
Enter the engineers
The latest announcement is the next step in meeting customer needs. “What we’ve increasingly realised is that you can take a strategist, you can take a scientist, but now you also need an engineer,” he explained.
The current bottleneck comes when agents have to be engineered into a company’s existing fabric of workflows and processes. The AWS Generative AI Innovation Centre is taking the strong foundation it has built over the past three years and moving into its next iteration, forward deploying cross-functional teams spanning science, engineering, strategy and security, he said.
The real test is if these FDEs can understand the business impact and reason for the investment into AI for the customer. If they understand this, customers are going to see significant benefits as a result
Peter Bryant, Omdia
An important aspect is the transfer of knowledge from these teams to customers so they can become self-sufficient, and a major part of this is helping customers adopt the AI-driven development lifecycle (AI-DLC), AWS’s AI-driven take on the traditional software development lifecycle.
“It’s not about adopting technology as much as it’s about how adopting AI changes how you operate as a team," he explained. The FDE teams drive the outcome of the project, and teach customers while doing so. “Part of our forward-deployed engineering strategy is enabling customer self-sufficiency.”
The centre is working with a number of Australian customers, particularly in sectors including financial services, energy and mining.
For example, the Commonwealth Bank of Australia (CBA) has created a technology hub in Seattle, supported by a dedicated team from the centre. The bank identifies potential use cases and sends appropriate groups of employees to Seattle, where they undertake three-week sprints to build and review solutions before returning home to put the results into production.
"We’ve been doing this now for over a year with the Commonwealth Bank, and it’s produced a number of solutions that have gone into production” in areas including business lending, risk management and core IT, said Rashid. “That’s one really good example of how we've taken the forward-deployed concept and found a neutral space to forward deploy both teams.”
Peter Bryant, global systems integrator practice leader at Omdia, said: “These FDE investments are coming ultimately because the current GTM [go-to-market approach] has not done enough to help customers derive value that is clear. Customers are going to gain access to some of the best engineers in the world directly helping them understand the best usage of the technology.”
He added: “The real test is if these FDEs can understand the business impact and reason for the investment into AI for the customer. If they understand this, customers are going to see significant benefits as a result. Partners will be critical to the long term of this. Two-thirds of all spend on technology flows through partners for a reason, which is that this is how customers prefer to buy. The ability of partners to support these customers is going to be the difference between success and failure when it comes to AI implementations in the long run.”
Forward-leaning Australia
Australian companies are generally well placed relative to their international peers. “I’ve been an executive sponsor to a number of our customers,” said Rashid. “When you look at capital-intensive industries such as energy and mining, which is obviously very big in Australia, I’m starting to see more of them wanting to leverage AI to solve many of their complex challenges.”
Those challenges include ageing systems and technical debt, with AI seen as a more capital-efficient way to modernise them.
“What I’ve been impressed with is how forward-leaning some of the companies in Australia are around this, and a large part of this has been driven by C-level leadership, whether it's the CFO, the COO or even the CEO really trying to drive that agenda," he said.
These projects address customer-facing issues, back-office issues or both. For example, a bank reworking its business lending practices with AI can improve the customer experience at the same time as improving throughput, so it can approve more loans and potentially enter new markets.
Turning to IT operations, some customers are looking to reduce the time it takes to discover all of their systems, so they can approach modernisation in a more thoughtful and systematic way.
“We’re actually working very closely with IT teams in using AI agents to do discovery of these systems, and using a technology called context graphs to be able to guide customers on their approach to modernisation,” he said.
In some cases, this involves legacy systems based on technologies such as Oracle and Teradata that have been running for decades. According to Rashid, there is a lot of tribal knowledge about such systems, but it is possible to use AI to first understand and then modernise them.
At the other extreme, “when we work with unicorn startups like Canva or publicly traded companies like Atlassian, we are doing very fine-grained science work with them related to models, and truly helping them get value out of their investments in AI”, he said.
Organising for AI
According to Rashid, it is very important to have a top-level strategy that not only the C-level leaders but also those below them subscribe to. The organisations that are able to move fastest, he suggested, are those that adapt themselves organisationally to AI before they adopt the technology broadly.
“This is something we actually do help customers with: how do you organisationally architect yourself to adopt AI and therefore move at the speed that AI can give?"
One of the practical ways the Generative AI Innovation Centre helps is through a programme called the Centre of Acceleration, which identifies the central components that need to be enabled across an organisation to provide consistency around use case identification and prioritisation, data connectivity, data governance, financial governance, responsible AI, observability and coding tools. Centralising these foundations accelerates the distributed teams building AI applications for their business units.
AWS’s own research into the Australian market showed 61% of businesses have adopted AI, so there is clearly a lot of ambition and growth in terms of wanting to use the technology, and the company wants to encourage structured and organised experimentation, Rashid said.
These technologies have low barriers to entry, which is a great equaliser when it comes to developers and other workers gaining access. Innovation and access are good things, but it is very important to have the right governance and structures in place, he explained.
“Cloud went through a very similar pattern. It was very accessible and teams started to use it, and then it got to a point where many teams were using it and there was a need for governance,” he said. “So, we’re leading with that story and saying: ‘Access to this technology is fully available, but how do we ensure that we get ahead of governance and operational controls so that you can be more structured about experimentation and be very intentional about how you allocate budgets towards various projects?’”
