Enterprise
AWS Forward Deployed Engineering Expands to 15 Countries
Amazon's AI implementation service now helps enterprises bridge the gap between prototype and production with embedded engineering teams and production-ready accelerators.
Key takeaways
- AWS Forward Deployed Engineering is now available in 15 countries, embedding AWS engineers with customers to build production AI systems in 45-day sprints.
- The service addresses three critical layers: building a shared foundation with governance, automating rule-heavy workflows, and implementing adaptive governance that scales.
- Samsung Life Insurance built a governed agentic platform in 45 days, while Marsh achieved up to 70 percent efficiency gains in targeted processes within two months.
- Customers own all work products, which run in their AWS accounts under their security controls, with their engineers trained to extend solutions independently.
- Each sprint has a set price billed through AWS accounts, with work beginning within days after initial setup by AI agents.
Amazon Web Services has expanded its Forward Deployed Engineering service to 15 countries, embedding AWS engineers directly with enterprise customers to move AI systems from prototype to production-scale deployment.
The service, which launched in June, now operates across the U.S., Mexico, Brazil, Canada, Australia, Singapore, India, South Korea, Japan, UK, Germany, Netherlands, France, Sweden, and Saudi Arabia for both commercial and public sector organizations.
Why it matters
Most enterprises have built AI prototypes that work in controlled environments but struggle to deploy them at production scale. The gap between demo and deployment—where issues like data quality, audit requirements, and undocumented business rules emerge—is where most AI initiatives stall. AWS's approach addresses this by embedding engineers who work in the customer's own AWS account, building production systems from day one rather than prototypes that need rebuilding.
Three-layer implementation framework
According to Dr. Asa Kalavade, who leads AWS Forward Deployed Engineering, the hardest problems appear in three distinct layers. First is the foundation layer: identity, security, context, governance, and operational controls that must be production-ready from the start and shared across use cases. The biggest gap here is typically context—most organizations have data spread across systems with no common ontology that AI agents need to reason effectively.
The second layer involves automating high-volume, rule-heavy workflows like claims processing, KYC reviews, and invoice reconciliation. The challenge isn't technical automation but uncovering business rules that live in employees' heads rather than documentation. Samsung Life Insurance used the service to build a governed agentic platform in 45 days, consolidating four separate affiliate efforts into a single control plane.
The third layer is governance that adapts as models, business rules, and regulations change. Every agent needs its own identity, permissions, and audit trail. Marsh, a global professional services firm, built controls in from the start and reached production in about two months, achieving efficiency gains up to 70 percent in targeted areas.
How the service works
AWS pairs a small team of engineers, scientists, and product managers—many from teams that build Amazon Bedrock and other AWS AI services—with customer engineers. They work in 45-day sprints, building directly in the customer's AWS account on their data and through their existing security controls.
The team deploys what AWS calls an "opinionated stack" that includes a living context of how the business works, a workbench for building, a runtime on Amazon Bedrock AgentCore, and a control tower for governance. Production-ready accelerators handle common tasks like document processing, allowing each new workload to focus on business-specific requirements.
Rio Tinto Iron Ore used the service to map a 30-year-old Oracle system with over 5,000 database tables—a process that previously took months but was completed in weeks using AI agents.
Customers own everything built during engagements. All work runs in their AWS account under their security controls, and their engineers work alongside AWS teams from the first sprint. When AWS engineers step back, customer teams can run and extend solutions independently.
Each 45-day sprint has a set price and is billed through the customer's AWS account. Organizations can start by contacting their AWS account team with an executive who owns the outcome and engineers ready to build.
These details were first reported by AWS in a blog post by Dr. Asa Kalavade.