Trane 如何借助 Amazon Bedrock AgentCore 将建筑洞察速度提升 60 倍

Trane Technologies 在全球管理数百万台互联的供暖、通风与空调（HVAC）资产，但要获得一个单一的运营答案，可能意味着需要交叉核对多个仪表板，并在菜单中逐层深入操作 20 分钟甚至更久。对于以这一规模运营的组织而言，这类摩擦会拖慢运营、延后纠正措施，并在整个企业范围内带来实质性的业务影响。

在 3—4 周内，Trane 的工程团队基于 Amazon Bedrock AgentCore 构建了一个由 AI 驱动的智能体解决方案，将原本需要 20 分钟、涉及多个屏幕的诊断工作流，缩短为一次仅需 20 秒的自然语言交互。这一结果基于 Trane 与技术人员开展数周内部基准测试所得。它代表着洞察获取速度提升 60 倍，帮助运营模式从被动响应转向更主动、数据驱动的优化。

本文介绍该解决方案背后的架构方法和关键设计决策：

• 将智能体逻辑与工具执行分离。
• 通过集中式工具网关集成实时遥测数据。
• 为不同角色定制响应。

Trane Technologies 与建筑智能挑战

Trane Technologies 是一家全球气候创新企业，年收入超过 210 亿美元，业务覆盖 100 多个国家。通过其战略品牌 Trane，该公司管理数百万台互联 HVAC 资产，覆盖数据中心、医院、制造设施及商业地产组合。这一庞大体系的核心是 Trane Cloud，这是一个汇集数百万 HVAC 系统实时性能数据的数字中枢。Trane Cloud 将原始设备遥测数据转化为可操作的智能信息，用于预测性维护、能源优化和卓越运营。

虽然基于仪表板的建筑管理系统为监控和控制提供了基础，但提取跨系统洞察仍可能要求用户在多个界面、分层菜单和彼此割裂的仪表板之间操作。通过将自然语言处理与 Trane Cloud 深度集成，用户可以通过单一的对话式界面获取这些运营上下文。结果是，用户能够更快获得复杂建筑管理问题的答案，并以更加主动、信息更充分的方式开展设施运营。

业务挑战：建筑运营人员为何需要 AI Agent

建筑运营人员、现场技术人员和服务经理手中拥有丰富的数据。来自不同系统的设备遥测数据、性能分析、故障警报、能耗模式和优化机会源源不断，但从中提取可操作洞察依然困难。根本问题在于，不同利益相关方对同一份数据需要截然不同的视图。

现场技术人员需要诊断精度。他们需要制冷剂压力、故障代码和系统级故障排除工作流。客户经理需要战略智能。他们需要正常运行时间指标、成本节约机会和资产组合性能趋势。建筑业主则要求高管层面的清晰信息。他们需要能效评分、可持续性指标和简化的运营摘要。现有工具为所有角色提供单一界面，要求每位用户都要浏览不属于其工作流的功能。

现有建筑管理应用依赖逐屏导航，这使跨设备比较更具挑战性，并要求用户记住菜单层级和技术术语。即使是基础的资产组合级问题，也可能需要耗时的人工工作流，并跨越多个屏幕完成。

Trane 的智能体式 AI 解决方案与 Trane Cloud 共同提供四项能力：

• 基于角色的访问控制，根据每位用户的权限和需求定制响应。
• 实时 HVAC 分析，即时访问当前和历史性能数据。
• 智能搜索，覆盖 Trane 的技术文档和最佳实践知识库。
• 可扩展架构，能够随 AI 能力的进步而演进，并扩展至更多建筑系统。

这些能力为跨角色、工作流和运营环境获取建筑智能信息，创造了一种更具可扩展性的方式。

解决方案概览：架构

扩展智能建筑运营所面临的挑战，在于将数百万台互联资产生成的海量数据转化为可操作洞察。尽管 Trane Cloud 能够大规模接收实时遥测数据，回答运营问题过去仍要求用户浏览彼此割裂的仪表板，并手动关联不同系统中的信息。为此，团队基于 Amazon Bedrock AgentCore 和 Strands framework 构建了一个对话式 Agent，并通过 AWS Cloud Development Kit (AWS CDK) 的基础设施即代码进行部署。团队选择 Strands 作为 Agent 框架层，是因为它提供了用于构建 Agent 行为的开发者 SDK 和编排逻辑；而 AgentCore 则负责底层托管运行时、记忆、工具网关和生产基础设施。

为避免单体设计的局限，该解决方案采用多智能体架构，每个专用助手均由各自的系统提示词约束，使其专注于单一能力领域：

• 资源助手（Resources Assistant）：检索并汇总参考资料。例如，用户可能会问：“我该联系谁、应遵循哪份手册，或可以向客户分享哪些文档？”
• 知识助手（Knowledge Assistant）：综合回答有关设备工作方式、系统参数，或 Trane Cloud 基础设施是否通过 SOC 2 鉴证等技术问题。
• 分析洞察助手（Analytics Insights Assistant）：查询实时遥测数据，发现能效提升机会、标记需要检查的项目，并追溯故障根因。
• 专家顾问（Expert Advisor）：帮助用户判断何种产品解决方案适合某个场景、如何最大化客户价值，或如何组装定制化演示。
• 导航助手（Navigation Assistant）：返回用户需要的准确链接和工具，涵盖技术支持升级表单到替换零部件订购页面。

该架构专为可扩展性而设计。团队可以通过 Amazon Bedrock AgentCore 的一项能力 AgentCore Gateway，以及模型上下文协议（Model Context Protocol，MCP）等开放标准，连接更多 Agent 或工具，例如工单管理系统和企业客户关系管理（CRM）系统。

图 1：Trane 在 Amazon Bedrock AgentCore 上构建的对话式 Agent 高层架构

微服务架构：系统设计与实施挑战

要在企业规模下支持这些差异化用户需求，需要一种能够让各项能力独立演进的架构。单体 Agent 会迫使所有变更——新工具、更新后的提示词、额外数据源——都经过同一条部署流水线，随着系统增长而形成瓶颈。

将 20 分钟的人工诊断缩短至 20 秒的对话，暴露出四项架构挑战。第一项挑战是集成。该解决方案必须将实时遥测数据与针对庞大知识库的智能搜索相结合，同时支持连接 CRM 等外部系统。第二项是分离。Agent 逻辑必须与后端工具执行解耦，以便二者可以独立部署，并拥有明确的职责边界。第三项是上下文。系统需要在故障排除会话中维持对话状态，同时避免搭建复杂的自定义向量数据库基础设施。第四项是可观测性。当一个 Agent 在多步骤推理链中编排多个工具时，故障会变得难以定位。错误答案可能源于缺失的 API 凭证、格式错误的工具响应，或模型幻觉。没有端到端追踪，团队便无法在生产规模下区分这些问题。

Amazon Bedrock AgentCore 如何应对这些挑战

Amazon Bedrock AgentCore 是一个智能体平台，可使用任意框架或模型，大规模构建、连接和优化 Agent。Trane 团队使用了四项 AgentCore 能力来应对前述挑战。

在选择 AgentCore 前，团队曾评估在 Amazon Elastic Container Service（Amazon ECS）和 AWS Lambda 上托管 Agent。这种方式将要求在计算层之上构建会话隔离、自动扩缩容逻辑以及按会话计费能力。四项差异化因素促成了最终决策。第一，托管 Agent 运行时减轻了基础设施运维负担，无需预置或扩容集群，也无需为用户会话之间的闲置容量付费。第二，内置会话记忆免除了搭建和维护外部向量数据库，或编写自定义上下文窗口管理代码的需求。第三，通过 AgentCore Gateway 进行的原生工具编排，可将现有内部 API 转换为与 Agent 兼容的工具，而无需为每个 API 编写自定义集成逻辑。第四，AgentCore 的框架无关设计意味着团队可以使用 Strands SDK，而不会被锁定在专有编排层中，从而在需求变化时保留灵活性。

• AgentCore Runtime：Trane 使用 Amazon Bedrock AgentCore 的一项能力 AgentCore Runtime，在具备独立 CPU、内存和文件系统的专用 microVM 中隔离每个用户会话。AgentCore Runtime 会在会话结束时终止并清理每个 microVM。团队选择它，是因为 microVM 模型将面向用户的 Agent 与后端 MCP Server 分离，使二者能够独立部署，并拥有明确的职责边界。AgentCore Runtime 从物理上隔离现场技术人员的会话与建筑业主的会话，无需自定义基础设施即可强化基于角色的访问控制。Trane 仅为会话期间的活跃计算付费。工具调用之间较长的暂停时间——这是智能体工作流的典型特征——不会累积成本。
• AgentCore Gateway：Trane 使用 AgentCore Gateway 将 Trane Cloud 的内部 API 公开为兼容 MCP 的工具，团队按能力领域对其组织，并与 Amazon OpenSearch Service 集成。团队选择它，是因为要将这一组助手连接到实时分析、设备遥测和问题诊断，需要一个能够处理身份验证和模式转换的统一访问层。未来的集成，如 CRM 和工单系统，可以通过同一 Gateway 连接，无需额外的集成工作。
• AgentCore Memory：Trane 使用 Amazon Bedrock AgentCore 的一项能力 AgentCore Memory，在故障排除会话中维持对话状态，使用户可以自然地进行追问（“现在把它与上个月对比一下”），而无须重新说明上下文。团队选择它，是因为替代方案是搭建独立的向量数据库，并编写自定义上下文窗口管理代码。AgentCore Memory 开箱即提供短期会话记忆，并具有 90 天到期生命周期，在上下文感知与存储效率间取得平衡，帮助 Trane 满足其数据保留政策。
• AgentCore Observability：Trane 使用 Amazon Bedrock AgentCore 的一项能力 AgentCore Observability，追踪 Agent 发起的工具调用，并定位多步骤推理链中的故障。团队是在开发期间遇到持续性工具故障后选择它的；若无端到端可见性，这些故障无法被识别。工程师通过 Amazon CloudWatch 中的 Agent 追踪，确认了多次调用中相同的失败模式，并将其追溯至 AWS Secrets Manager 中缺失的密钥。添加该密钥后，故障得到解决。

实时数据流的工作方式如下：

• 携带 JSON Web Token（JWT）的用户查询进入 AgentCore Runtime。Runtime 的入站授权器（Amazon Bedrock AgentCore 的一项能力 AgentCore Identity）会根据 OpenID Connect（OIDC）发现端点验证该令牌。随后，AgentCore Memory 会在 Agent 处理请求前注入此前的对话上下文。
• 然后，Agent 通过 OAuth 2.0 机器到机器身份验证触发独立的 MCP Server，以安全访问后端工具。
• MCP Server 并行调用 OpenSearch Service 和其他已连接系统中的工具，实现快速的语义数据检索。
• 最后，Amazon Bedrock 上的 Anthropic Claude 模型综合遥测数据，并通过 Server-Sent Events（SSE）将响应流式传回用户。这可降低感知延迟，在数秒内提供答案。有关不同 AWS 区域的模型可用性，请参阅 Amazon Bedrock 中按 AWS 区域列出的支持模型。

成果与影响

通过使用 Amazon Bedrock AgentCore 和 Strands framework，团队实现了前文所述的洞察获取速度提升 60 倍。过去需要在多个屏幕之间导航才能完成的跨设备比较和诊断工作流，如今可通过一次自然语言查询在数秒内完成。这有助于团队更快行动、减少分析步骤，并以更加主动的方式开展设施运营。

除运行时性能外，该架构还带来了额外收益：

• 从原型到生产的快速推进： 团队 st

Trane Technologies manages millions of connected heating, ventilation, and air conditioning (HVAC) assets worldwide, but getting a single operational answer could mean cross-referencing multiple dashboards and drilling through menus for 20 minutes or more. For organizations operating at this scale, that kind of friction slows operations, defers corrective action, and creates material business impact across the enterprise.

In 3–4 weeks, Trane’s engineering team built an AI-powered agentic solution on Amazon Bedrock AgentCore that reduced a 20-minute multi-screen diagnostic workflow to a 20-second natural language interaction. This is based on Trane’s internal benchmarking with technicians over several weeks. This represents a 60x improvement in time-to-insight, helping shift operations from reactive response to more proactive, data-driven optimization.

In this post, we describe the architectural approach and key design decisions behind the solution:

• Separating agent logic from tool execution.
• Integrating real-time telemetry through a centralized tool gateway.
• Tailoring responses to different personas.

Trane Technologies and the building intelligence challenge

Trane Technologies is a global climate innovator with over $21 billion in annual revenue and operations in more than 100 countries. Through its strategic brand Trane, the company manages millions of connected HVAC assets, spanning data centers, hospitals, manufacturing facilities, and commercial real estate portfolios. At the heart of this vast landscape is Trane Cloud, a digital hub that aggregates real-time performance data from millions of HVAC systems. Trane Cloud transforms raw equipment telemetry into actionable intelligence for predictive maintenance, energy optimization, and operational excellence.

While dashboard-based building management systems provide a foundation for monitoring and control, extracting cross-system insights can still require users to navigate multiple screens, layered menus, and disconnected dashboards. By combining natural language processing with deep integration into Trane Cloud, users can access that operational context through a single conversational interface. The result is faster answers to complex building management questions and a more proactive, informed approach to facility operations.

Business challenge: Why building operators need an AI agent

Building operators, field technicians, and service managers have abundant data at their fingertips. Equipment telemetry, performance analytics, fault alerts, energy consumption patterns, and optimization opportunities flood in from disparate systems, yet extracting actionable insights remains difficult. The fundamental problem is that different stakeholders need radically different views of the same data.

Field technicians require diagnostic precision. They need refrigerant pressures, fault codes, and system-level troubleshooting workflows. Account managers need strategic intelligence. They need uptime metrics, cost savings opportunities, and portfolio performance trends. Building owners demand executive clarity. They need efficiency scores, sustainability metrics, and simplified operational summaries. Existing tools present a single interface across all roles, requiring each user to navigate features outside their workflow.

Existing building management applications rely on screen-by-screen navigation that makes cross-equipment comparison more challenging and demands that users memorize menu hierarchies and technical terminology. Even basic portfolio-level questions can require time-intensive manual workflows across multiple screens.

Together, Trane’s agentic AI solution and Trane Cloud deliver four capabilities:

• Role-based access control that tailors responses to each user’s permissions and needs.
• Real-time HVAC analytics providing instant access to current and historical performance data.
• Intelligent search across Trane’s knowledge base of technical documentation and best practices.
• Extensible architecture that evolves with advancing AI capabilities and expands to additional building systems.

These capabilities create a more scalable way to access building intelligence across roles, workflows, and operational environments.

Solution overview: Architecture

The challenge of scaling intelligent building operations lies in turning the massive volume of data generated by millions of connected assets into actionable insight. Although Trane Cloud ingests real-time telemetry at scale, answering operational questions has traditionally required users to navigate disconnected dashboards and manually connect information across systems. To solve this, the team built a conversational agent on Amazon Bedrock AgentCore and the Strands framework, deployed through AWS Cloud Development Kit (AWS CDK) infrastructure as code. The team chose Strands for the agent framework layer because it provides the developer SDK and orchestration logic for building agent behavior, while AgentCore handles the managed runtime, memory, tool gateway, and production infrastructure underneath.

To avoid the limitations of a monolithic design, the solution uses a multi-agent architecture where each specialized assistant is governed by its own system prompt, keeping it tightly focused on a single capability domain:

• Resources Assistant – Retrieves and summarizes reference material. For example, a user might ask: “Who do I contact, what manual should I follow, or what documentation can I share with the customer?”
• Knowledge Assistant – Synthesizes technical answers about how equipment works, its system parameters, or whether Trane Cloud’s infrastructure is SOC 2 attested.
• Analytics Insights Assistant – Interrogates live telemetry to surface efficiency opportunities, flag items needing inspection, and trace fault root causes.
• Expert Advisor – Helps users decide which product solution fits a scenario, how to maximize customer value, or how to assemble a customized demo.
• Navigation Assistant – Returns the exact links and tools a user needs, from the tech support escalation form to the replacement-parts order page.

This architecture is designed for extensibility. Teams can connect additional agents or tools, such as work order management systems and enterprise customer relationship management (CRM) systems, through AgentCore Gateway, a capability of Amazon Bedrock AgentCore, and open standards like the Model Context Protocol (MCP).

Figure 1: High-level architecture of Trane’s conversational agent on Amazon Bedrock AgentCore

Microservices architecture: System design and implementation challenges

Supporting these distinct user needs at enterprise scale requires an architecture that can evolve independently across capabilities. A monolithic agent would force every change (new tools, updated prompts, additional data sources) through a single deployment pipeline, creating bottlenecks as the system grows.

Reducing a 20-minute manual diagnosis to a 20-second conversation surfaced four architectural challenges. The first challenge was integration. The solution had to combine real-time telemetry with intelligent search across an extensive knowledge base while supporting connections to external systems like CRMs. The second was separation. Agent logic had to be untangled from backend tool execution so the two could deploy independently, with clear ownership boundaries. Third was context. The system needed to maintain conversational state across troubleshooting sessions without standing up complex custom vector database infrastructure. Fourth was observability. When an agent orchestrates multiple tools across a multi-step reasoning chain, failures become difficult to localize. A wrong answer could stem from a missing API credential, a malformed tool response, or a model hallucination. Without end-to-end tracing, the team had no way to distinguish between them at production scale.

How Amazon Bedrock AgentCore addresses the challenges

Amazon Bedrock AgentCore is an agentic platform to build, connect, and optimize agents at scale, with any framework or model. The Trane team used four AgentCore capabilities to address the preceding challenges.

Before selecting AgentCore, the team evaluated hosting the agent on Amazon Elastic Container Service (Amazon ECS) and AWS Lambda. That approach would have required building session isolation, auto scaling logic, and per-session billing on top of the compute layer. Four differentiators drove the decision. First, the managed agent runtime alleviates infrastructure operations. There are no clusters to provision or scale, and no idle capacity to pay for between user sessions. Second, built-in session memory removes the need to stand up and maintain external vector databases or build custom context-window management code. Third, native tool orchestration through AgentCore Gateway turns existing internal APIs into agent-compatible tools without writing custom integration logic for each one. Fourth, AgentCore’s framework-agnostic design meant the team could use the Strands SDK without being locked into a proprietary orchestration layer, preserving flexibility as requirements change.

• AgentCore runtime : Trane uses AgentCore runtime, a capability of Amazon Bedrock AgentCore, to isolate each user session in a dedicated microVM with its own CPU, memory, and filesystem. AgentCore runtime terminates and sanitizes each microVM on session completion. The team chose it because the microVM model separates the user-facing agent from the backend MCP Server, letting the two deploy independently with clear ownership boundaries. AgentCore runtime physically isolates a field technician’s session from a building owner’s, reinforcing role-based access without custom infrastructure. Trane pays only for active compute during a session. The long pauses between tool calls (typical of agentic workflows) don’t accumulate cost.
• AgentCore Gateway : Trane uses AgentCore Gateway to expose Trane Cloud’s internal APIs as MCP-compatible tools that the team organized by capability domain and integrated with Amazon OpenSearch Service. The team chose it because connecting the suite of assistants to real-time analytics, equipment telemetry, and issue diagnostics required a single access layer that handles authentication and schema translation. Future integrations (CRMs, work order systems) connect through the same Gateway without additional plumbing.
• AgentCore memory : Trane uses AgentCore memory, a capability of Amazon Bedrock AgentCore, to maintain conversational state across troubleshooting sessions so users can ask natural follow-ups (“now compare that to last month”) without re-specifying context. The team chose it because the alternative was standing up a separate vector database and writing custom context-window management code. AgentCore memory provides short-term session memory out of the box, with a 90-day expiry lifecycle that balances contextual awareness with storage efficiency, helping Trane address their data retention policies.
• AgentCore Observability : Trane uses AgentCore Observability, a capability of Amazon Bedrock AgentCore, to trace tool calls an agent makes and isolate failures across multi-step reasoning chains. The team chose it after encountering consistent tool failures during development that could not be identified without end-to-end visibility. Through the agent traces in Amazon CloudWatch, engineers confirmed the identical failure pattern across multiple invocations and traced it to a missing secret in AWS Secrets Manager. After adding the secret, the failures resolved.

The real-time data flow works as follows:

• The user’s query, carrying a JSON Web Token (JWT), hits AgentCore runtime. The Runtime’s inbound authorizer (AgentCore Identity, a capability of Amazon Bedrock AgentCore) validates the token against the OpenID Connect (OIDC) discovery endpoint. AgentCore memory then injects previous conversation context before the agent processes the request.
• The agent then triggers the separate MCP Server through OAuth 2.0 machine-to-machine authentication to securely access backend tools.
• The MCP Server executes parallel tool calls to OpenSearch Service and other connected systems for rapid semantic data retrieval.
• Finally, Anthropic Claude models on Amazon Bedrock synthesize the telemetry and stream the response back to the user through Server-Sent Events (SSE). This minimizes perceived latency, delivering answers in seconds. For model availability by AWS Region, refer to Supported models by AWS Region in Amazon Bedrock.

Results and impact

Using Amazon Bedrock AgentCore and the Strands framework, the team achieved the 60x time-to-insight improvement described earlier. Cross-equipment comparisons and diagnostic workflows that once required navigating multiple screens now complete through a single natural language query in seconds. This helps teams move faster, reduce analysis steps, and take a more proactive approach to facility operations.

Beyond runtime performance, the architecture delivered additional benefits:

• Rapid prototyping to production: The team stood up a working prototype and demoed it to stakeholders within 3–4 weeks, then rolled the agent out first to internal field technicians as beta testers. Field technicians exercise the most demanding diagnostic workflows, so their feedback surfaced accuracy and usability gaps under real conditions. The team refined the agent responses based on that feedback before releasing to external customers.
• Rapid integration across applications: With decoupled microservices architecture and the centralized AgentCore Gateway, integrating the same agent backend into other enterprise applications took less than a day.
• Effective production debugging: The full invocation traces from AgentCore Observability in CloudWatch let the team pinpoint a recurring tool failure quickly.

Production controls and responsible AI

Deploying an AI agent that returns diagnostic recommendations from live operational data requires controls against misuse and out-of-scope responses. The team configured Amazon Bedrock Guardrails with content filters to block harmful or inappropriate outputs and topic denial policies that restrict the agent to its operational domain, helping restrict it from answering questions outside its operational domain. Sensitive information filters detect and redact personally identifiable information (PII) before responses reach the user, and prompt attack detection helps guard against jailbreak attempts that could bypass the agent’s system prompt.

At the infrastructure layer, Trane’s role-based access model helps restrict each user to data within their authorization scope. AgentCore runtime’s per-session microVM isolation helps prevent cross-tenant data leakage risk at the compute level. These controls work together so Trane can scale the agent to production users with confidence. Responses stay within scope, role-based access and PII filters help protect sensitive data, and Bedrock Guardrails help enforce the agent’s intended operational boundaries.

Conclusion

Trane’s implementation demonstrates how organizations can use AI agents and their proprietary data as a distinct competitive advantage. By building on Amazon Bedrock AgentCore, the organization established a robust, secure enterprise standard that can be scaled and reused across different business units.

This phased rollout, from a 3–4 week prototype, through accuracy tuning and an internal beta, to external release and reuse, surfaced five insights for scaling enterprise AI:

• Data is your differentiator : The foundation of the agent’s success relies on existing enterprise data, such as OpenSearch Service indexes and Internet of Things (IoT) telemetry. The agent’s true value comes directly from using this proprietary data to drive more efficient energy usage and operational excellence.
• Architectural foundation matters: The Strands framework helped the team get started and iterate rapidly, while Amazon Bedrock AgentCore provided the purpose-built infrastructure and secure services necessary for running agents at scale. Committing to open standards like MCP helped provide flexibility for future enhancements and accelerated the build.
• Phase your rollout, harden internally first: Deploying to internal field technicians (the most demanding users) captured critical technical feedback and let the team refine accuracy before releasing externally to Trane Cloud customers.
• Role-based access and responses: Different roles require different tools and different response formats. Dynamic policy mapping means a field technician receives step-by-step diagnostic workflows while a building owner receives simplified efficiency scores. This drives adoption because users see responses tailored to their role.
• Observability is key for agentic systems: When an agent orchestrates multiple tools in a reasoning chain, failures can be hard to identify. End-to-end invocation tracing through CloudWatch gave the team the ability to pinpoint and resolve issues quickly.

To learn more about building agentic solutions, visit the Amazon Bedrock AgentCore documentation, explore the Strands Agents SDK, or follow the AWS Artificial Intelligence Blog for more customer stories.

What’s next

The team is expanding the solution in several areas.

First, AgentCore Evaluations, a capability of Amazon Bedrock AgentCore, will gate future rollout phases on automated accuracy thresholds. Easing the burden on the quality assurance (QA) team, the team will define pass-rate benchmarks that must clear before each release moves forward.

Second, Policy in Amazon Bedrock AgentCore will replace custom authorization logic currently coded into the Runtime. Tool-level access decisions will shift to declarative policy definitions, making it easier to onboard new roles and audit who can invoke which tools.

Third, the team is opening their AgentCore Gateway to other engineering teams across Trane. Because the Gateway exposes tools through MCP, an open standard, other teams can build their own agents on top of the same centralized data layer without learning proprietary interfaces or standing up duplicate integrations. The team is building self-service onboarding with usage tracking so they can measure adoption and identify which tools other teams find most valuable.

Fourth, the team is adding new tools at a rapid pace. One example already in progress: field offices currently perform deep cost savings analyses manually for each customer site. The team is building this as an agent tool so the analysis runs on demand through natural language, removing hours of manual work per engagement.

---

About the authors

Senthil Chinnaiyan

Senthil, Director of Engineering – Digital, PGT (Product Growth Team) and AWS Solutions Architect at Trane Technologies, leads a high-performing engineering organization building highly scalable, cloud-native HVAC solutions that process billions of data points daily. With 29 years of experience bridging technology and business outcomes, he is spearheading Trane’s commercial HVAC transformation toward Agentic AI, delivering end-to-end intelligent systems that integrate Amazon Bedrock, AgentCore, and broader AWS services to drive measurable business value. Senthil is known for executing with relentless urgency, architecting solutions that are highly scalable, and consistently achieving cost-optimal results that compound into strategic enterprise advantage.

Subbha Praveen Tadisetty

Subbha is a Digital System Architect at Trane Technologies with over 21 years of experience leading high-impact enterprise architecture, cloud modernization, and AI-driven integration across digital platforms. An AWS Certified AI Practitioner and AWS Solutions Architect, he is passionate about building scalable serverless systems, automating deployment pipelines, and developing agentic AI solutions. Based in White Bear Lake, Minnesota, Praveen enjoys staying at the forefront of emerging cloud technologies and helping engineering teams drive continuous innovation.

Alex Jones

Alex is an Application Architect with Trane Technologies working on user management and AI features for the Trane Cloud platform. His primary areas of work focus on building scalable backend solutions on AWS that enable critical business processes. In his personal life, Alex lives in Minneapolis and enjoys cooking and tinkering in his home lab.

Dave Shimko

Dave is a Solutions Architect with Amazon Web Services (AWS), working with Automotive and Manufacturing customers to accelerate their cloud and AI journey. He is passionate about helping customers adopt containers, generative AI, and modern developer experiences. Outside of work, Dave lives in Raleigh, North Carolina and enjoys traveling, being outdoors, and spending time with his family.

KP Babu

KP is a Senior GenAI/ML Specialist Solutions Architect at AWS, where he helps enterprises build generative AI systems that pair cutting-edge innovation with responsible deployment and meet stringent security and compliance requirements. Drawing on extensive experience across a wide range of customer engagements, he focuses on designing agentic AI systems and orchestration frameworks that streamline complex workflows while preserving appropriate human oversight. Through his technical publications and speaking engagements, he translates hard implementation challenges into practical guidance, always coming back to the same point: successful AI agents earn their place by solving genuine business problems through effective reasoning, planning, and action within well-defined enterprise contexts.

Will Krinickas

Will is a Senior Generative AI Specialist at AWS, bringing over a decade of experience in Data & AI. In his role, he drives GenAI adoption across the Automotive & Manufacturing (AutoMfg) industry. He partners with enterprise customers to move generative AI workloads from experimentation into production, pairing deep technical fluency across the GenAI stack, industry insights, and a work-backwards approach rooted in customers’ business goals.