跳到正文
原文
AWS Machine Learning Blog· Dr. Asa Kalavade·· 4 小时前

AWS Forward Deployed Engineering 扩展至 15 国,助力企业 AI 落地生产

Getting AI out of the demo and into businesses globally

SI 导读

AWS Forward Deployed Engineering 现已在美国、日本等 15 个国家面向商业及公共部门客户开放,旨在解决 AI 从演示到生产的“最后一公里”难题。该服务通过构建基础架构、自动化工作流及治理体系,帮助三星生命保险在 45 天内建立受治理的智能体平台,并协助 Marsh 实现高达 70% 的效率提升。AWS 于 6 月推出该服务,以填补原型与生产环境间的差距。

SI 评分32
正文

AWS Forward Deployed Engineering is now available in 15 countries including U.S., Mexico, Brazil, Canada, Australia, Singapore, India, South Korea, Japan, UK, Germany, Netherlands, France, Sweden, and Saudi Arabia for commercial and public sector customers; and powered by dozens of built in production-ready accelerators.

Every enterprise I talk to has built something with AI that works. A prototype that reads a claim. An agent that resolves a card dispute. A model that answers a customer question in seconds.

The prototype works. The demo impresses stakeholders. Then the organization tries to put it into production, and the real problems begin. Production data is messier than the test set. Auditors want to know why the agent made each decision. Volume grows from a pilot group to the whole company. The rules that matter most turn out to live in people’s heads.

None of these are model problems. They’re business problems, and they decide whether AI ever pays off. The stretch from demo to production is the last mile, where most AI stalls. It’s also where our teams spend their time.

We launched AWS Forward Deployed Engineering (FDE) in June to close that gap. Since then, we’ve worked through these problems with commercial and public sector customers in financial services, healthcare, sports, media, retail, and more. Here’s what we’ve learned, and how it shaped what we built.

The problem has three layers

Model providers, cloud providers, and consulting firms are all embedding engineers with customers. At AWS FDE, we measure success by what remains after our engineers leave: a foundation your team owns, workflows that keep learning, and governance you can trust.

Across customers, the hardest problems show up in three layers: the foundation, the workflows, and governance. Each one is harder than it looks, and each one builds on the one before it. That holds whether you’re starting fresh or building on existing systems. On a greenfield build, you can move fast, but there’s no foundation to stand on yet. On a brownfield system, agents must work within decades of data, rules, and integrations.

Layer one: A foundation that compounds

Before an agent can do useful work at scale, it needs a foundation: identity, security, context, governance, and operational controls. That foundation must be production-ready from day one and shared by every use case that follows. Context is often the biggest gap. Most customers have data spread across systems and no common ontology, the shared model of how the business works that agents need to reason over.

Without a shared foundation, the second pilot costs nearly as much as the first. Integrations get rebuilt, and lessons fail to carry forward. Most organizations reach a point where they stop and build the foundation first, so each new use case starts from what the last one built instead of from scratch. We call this the AI reset.

Samsung Life Insurance had four affiliates, each building its own AI point solutions from scratch. In 45 days, the FDE team built a governed agentic platform with a single control plane and compliance built in. New agents now deploy through registration and policy assignment instead of a rebuild.

Layer two: Automating workflows

Next is the work itself: High-volume, rule-heavy processes such as claims processing, know your customer (KYC) reviews, card disputes, loan origination, invoice reconciliation, permitting, and customer service. This is where AI can deliver immediate and measurable outcomes. The payoff is fast, but the knowledge that drives these processes rarely lives in a system. It lives with the people who do the work.

The hard part isn’t automation itself. It is uncovering the business rules behind each decision. When an invoice and a purchase order disagree, which record wins? When can an exception clear on its own, and when does it need a person? Experienced employees answer these questions from judgment, and that judgment is rarely written down.

Sometimes the knowledge is buried in the systems themselves.

Rio Tinto Iron Ore runs one of the world’s most complex mine-to-ship supply chains on a 30-year-old Oracle system. That system has more than 5,000 database tables and 140 applications. Mapping it had been a multi-month exercise. The FDE team built agents that mapped it in weeks, surfacing the dependencies that Rio Tinto’s teams could validate and build against.

Mapping what exists is only the start. The biggest opportunities emerge when organizations stop asking how AI can make existing work more efficient and start asking how it can change how work gets done.

Layer three: Governance that holds up as everything changes

Governance isn’t a phase of the project. It’s part of the architecture. As agents multiply, so do security risks, costs, and decisions you must explain. Every agent needs its own identity, permissions, and audit trail, separate from human identity. And because the models underneath are probabilistic, agents need monitoring that keeps their results predictable and auditable long after launch.

The system also must keep working as everything around it changes. Models drift, business rules shift, data formats change, and regulations tighten. New security issues appear, and code falls out of date. As newer models arrive, the instructions and tooling around them will change too. You need evaluations and guardrails that keep agents predictable where they must be and creative where they can be.

When governance comes at the end, it becomes the gate that stops production. When it’s built in from day one, it’s what lets your business move fast.

Marsh, a leading global professional services firm, needed agents that would hold up under its own audit standards. The FDE team built those controls in from the start, and the first use case reached production about two months after work began. Efficiency gains in the targeted areas have been as much as 70 percent, with some process steps moving from hours to seconds. Marsh has since expanded the work and keeps extending the system itself.

How we solve it: Speed, scale, and control

Our perspective comes from building and running these systems ourselves. AWS FDE pairs a small team of AWS engineers, scientists, and product managers with your engineers and a crew of specialized AI agents. Many of our FDEs come from the service teams that build Amazon Bedrock, Amazon Bedrock AgentCore, Kiro, AWS Quick, and AWS Security Hub. They’ve built and run AI products at AWS scale, and they bring those engineering practices into your environment. Together, our teams take a high-value business problem to production in 45-day sprints, in your own AWS account. Customers shouldn’t need to choose between speed, scale, and control. FDE is built to deliver all three.

Speed – You might come to us with a broad goal or an ambiguous problem. We refine it with your business stakeholders, then build in your environment from day one, on your data and through your identity and security controls. We build in short cycles alongside the people who do the work. Each cycle puts real cases in front of them, and the unwritten rules come out as they react. We capture those rules in a living context, so the system keeps learning after we step back. Nothing gets rebuilt between prototype and production. You see working demos throughout, evaluations track progress against the outcomes you care about, and you have a production system within a quarter or less.

Scale – A point solution can move fast on one use case but leaves no foundation for the next five. Every FDE team works the same way, using the AI-Driven Development Lifecycle (AI-DLC), the method our own service teams use to take a broad problem to a working outcome. We deploy an opinionated AWS stack that reflects what we’ve learned building AI services, adapted to your environment: a living context of how your business works, a workbench where your engineers and agent crews build, a runtime on Amazon Bedrock AgentCore, and a control tower for governance, outcomes, cost, and agent performance. Accelerators for common tasks, such as document processing and customer sentiment analysis, handle the standard pieces, so each new workload can focus on what’s specific to your business. The opinionated AWS stack, context, and controls carry from one use case to the next.

Your context is what makes scaling possible. Say your first use case is a sales automation workflow. Because its context lives in your AWS account, the next use case, supply chain optimization, builds on it instead of starting over. Your context, tools, and AWS services all sit in one place, so each use case is faster to build than the last.

Control – You own what we build. Everything runs in your AWS account under your identity, network, and security controls. Your engineers work alongside ours from the first sprint, learning both the architecture and the delivery approach on a real business problem. They can then run and extend the solution long after the engagement ends. When we step back, your team runs the system from a control tower, a single view to govern, secure, and monitor every agent. We measure success by how much your team can build without us.

After the first or second project, we want you running AI software factories of your own. One financial services customer did exactly that with AI-DLC, with agents and engineers building together. Throughput rose 53%, and code reviews got 56% faster.

You also keep your choices open. Models are improving fast. When a better or more cost-effective model arrives, you can adopt it through Amazon Bedrock without rebuilding, so you control both capability and cost. You can also bring your own coding agents. Our opinionated AWS stack, adapted to your environment, connects to your systems through APIs, open standards, and connectors, and it works in hybrid environments. You keep the tools and partners you already use.

A flywheel where everyone gets better

Your team gets better. Your capability grows with every use case. The living context, the opinionated AWS stack, and your team’s skills carry forward, so the second use case doesn’t rebuild what the first one built.

AWS Services get better. Our engineers bring what they learn from customers back into the products. You build on the latest services, not a custom solution that falls out of date.

All of AWS works around you. Customer development engineering explores early ideas, FDE takes the hardest problems to production, AWS Professional Services and AWS Partners extend the work across your organization, and AWS Support keeps it running. You get the full depth of AWS behind your AI, and you stay in control of how you use it.

What’s new, and how to start

AWS FDE is now available in 15 countries to commercial and public sector customers, and it’s simpler to buy. Each 45-day sprint has a set price under the FDE Terms and is billed through your AWS account.

We work with customers wherever they are in their transformation. For customers starting out, we build a secure, governed foundation and the first workflows on it. For those ready to change how work gets done, we rebuild workflows such as claims processing and KYC reviews, along with the ones their customers use directly, like shopping and issue resolution. Agents then run the work. For those held back by legacy applications, we work with their engineers to build software factories they can reuse for each new workflow. For those already scaling in production, we keep systems secure with agents that find and fix issues, and we lower the cost per outcome.

If you’re ready to start, talk with your AWS account team. Bring an executive who owns the outcome and engineers ready to build alongside ours. We’ve learned where projects stall before they start, so our agents take on much of the setup, like access, data, and security checks. Work can begin within days. Then we agree together on what the first sprint will deliver and how we’ll measure success.

The gap between AI that works and AI that changes business outcomes is real. It is an engineering, operational, and organizational one, and our teams help customers close it every day. The organizations that succeed with AI won’t necessarily have the most advanced models. They’ll be the ones that build the foundation, workflows, and governance to scale AI across the business.

We’re still early, and we learn something new with every customer. The models will continue to evolve, and so will the tools around them. What should last is a foundation you own, workflows that keep learning, and governance you can trust. That’s what we mean by what remains after our engineers leave.

To get started, contact your AWS account team or learn more about AWS FDE.


About the author

Dr. Asa Kalavade

Dr. Asa Kalavade

Asa is a technology and product leader at AWS who turns emerging technologies into scalable solutions for enterprise customers. She leads AWS Forward Deployed Engineering. Before that she led AWS Transform, which offers agentic AI based modernization of legacy software and infrastructure, and the transformation of AWS’s go-to-market tools and processes. In more than a decade at AWS, she has led seven 0-to-1 launches, built and scaled global teams, and led businesses through significant growth. Before AWS, Asa founded and ran two venture-backed startups, and she began her career at Bell Labs Research. She holds a Ph.D. in EECS from the University of California, Berkeley, and more than 40 patents.

来源:AWS Machine Learning Blog · aws.amazon.com

© 2026 SI·Hot · Super Intelligence Hot · 超级智能热点 · 网站数据均来源于网络公开资料,版权归来源方所有