How the agent works.
A secure path from your question to an answer grounded in approved Scania information and tools.
Technical details Identity, knowledge, files, and operations
Identity & edge protection
Entra ID validates the signed-in user. CloudFront and AWS WAF protect the site and APIs, and backend handlers reject requests that bypass the approved origin.
Knowledge & tools
Cloud Docs are retrieved through a managed AWS Knowledge Base and AgentCore gateway. Approved account lookups run through the backend without exposing AWS credentials to the browser.
Files & voice
Realtime dictation streams directly to Amazon Transcribe in eu-west-1 through a short-lived signed session. Review attachments use private temporary S3 storage.
Reliability & feedback
CloudWatch metrics, tracing, alarms, and usage telemetry help operate the service. Analytics can include identifiable question text and signed-in profile information; see Data & Privacy for details. Answer feedback is written through a protected API to a dedicated store.
What’s new.
The meaningful milestones as the agent evolved — new capabilities, quality improvements, and platform hardening.
Oct 2026 Native Confluence integration New
The agent now has a native Confluence integration and can answer from documentation kept in Confluence. This is a major addition: teams that document their services in Confluence can now make that documentation available to the agent with little effort. Integration Platforms is the first team to connect its documentation.
- Ask about API Manager and the Scania Streaming Platform (Kafka), including publishing and subscribing to APIs, producer and consumer onboarding, Entra ID access and API security guidelines.
- Integration Platforms is now an information area on the homepage.
- Confluence pages are listed under Sources with the Confluence mark and open the original page.
Oct 2026 Intelligent UI in chat New
Answers can now combine clear explanations with useful visual modules.
- Diagrams, compact tables, callouts, code, timelines, status cards, charts, checklists and icon-led lists help explain the answer when useful.
- Clearer borders, consistent text sizes and spacing, and connected diagram frames make visual answers easier to scan on desktop and mobile.
- Text and completed components appear progressively with soft fades. Reduced-motion preferences are respected.
- Diagrams and flows documented as pictures in the knowledge base can be redrawn as visual modules from their indexed descriptions.
Aug 2026 Expanded Cloud, Devtools and IT Security guidance New
More approved Scania information is now available to the agent, making it easier to get relevant answers across everyday cloud, development and security work.
- Cloud includes Cloud Docs, Cloud Learning Hub material, AWS account details, guardrails and approved architecture patterns.
- Devtools includes developer onboarding, Jira and GitLab access, developer best practices and security guidance.
- IT Security includes Secure IT guidance, security best practices and architecture review guidance.
- Architecture explanations and Cloud security reviews can use the expanded guidance for more relevant recommendations.
Aug 2026 Recent chats, ready to continue New
Recent conversations are now available from the sidebar, so you can return to earlier work and continue where you left off.
- Your 20 most recent conversations are available in Recents.
- Clear, automatically generated titles make conversations easier to find.
- Reopen full context, rename or delete individual chats, or clear the history.
Aug 2026 Cloud security review preview & easier sharing Preview
Cloud security review is now available in preview. You can upload architecture diagrams and PDFs for a review grounded in the AWS Well-Architected Security Pillar and relevant Scania cloud guidance. You can also share any question with others through a copied link.
- Add images or PDFs with the file picker, drag and drop, or clipboard paste.
- Continue with follow-up questions while the review context stays active.
- Review attachments expire automatically after 24 hours.
- Share a question with a copied link.
Jul 2026 Sharper answers & a livelier prompt New
The agent now runs on Amazon Bedrock, using Claude models selected for responsive chat and in-depth security reviews.
- Animated example prompts on the landing page to spark ideas.
- The agent can proactively offer a relevant follow-up question.
- Cleaner sources: citation markers no longer leak into answer text.
Jul 2026 Feedback & observability New
You can now rate answers with thumbs up or down and an optional comment, helping us improve the agent.
- New feedback capture with a dedicated store and API.
- Operations dashboard, tracing, and alarms for reliability.
- Usage analytics help us understand service adoption, common questions and reliability.
Jun 2026 Real-time voice dictation Improved
Dictation was upgraded to stream your words in real time as you speak, with a smoother reveal and preserved utterances.
- Manual edits to dictated text are now preserved.
- Faster, more reliable startup with a clear loading state.
Jun 2026 Refreshed Domains, Architecture & FAQ Improved
The landing page information sections were rewritten for clarity, with an updated architecture flow that explains how the agent works end to end.
- Clearer architecture diagram and Bedrock labelling.
- Polished source controls and a smooth scroll-to-bottom button.
Jun 2026 Hardened platform & user context Security
Agent prompts and model settings are now owned in code for consistency, and signed-in user context is passed to the agent for more relevant answers.
- Streaming chat moved into the VPC for a tighter network boundary.
- Self-hosted MSAL bundle and reduced backend log noise.
- Development-environment ribbon to avoid confusion between environments.
May 2026 Sign-in with Scania identity New
Entra ID authentication was added, so the agent greets you by name and shows your profile. A clean profile menu lives in the header.
- Personalised welcome greeting and profile photo.
- Appearance settings with light, dark, and system themes.
May 2026 Streaming chat experience New
Answers now stream token by token with a live activity view, a modern chat layout, and a stop control — plus multi-turn follow-up conversations.
- Streamed agent activity and thinking states.
- Copy and answer-detail controls on every response.
- Follow-up questions keep the conversation in context.
May 2026 Voice input & richer answers New
Added dictation voice input alongside typed prompts, and improved how answers render.
- Source links, tables, and correct list rendering.
- Collapsible Domains, Architecture, and FAQ info sections.
May 2026 First public preview Launch
The serverless infra AI agent went live at agent.scania.com, behind Scania’s network protections, with a calm single-prompt landing page.
- Custom domain with WAF-protected API origin.
- Suggested prompts and a polished answer experience.
Questions & answers.
Find answers about using the Infra Agent and how your information is handled.
Using the Infra Agent
Learn what the Infra Agent can help you with, how to get useful answers, and where to turn if you need more help.
What can the Infra Agent help me with?
The Infra Agent helps you find Scania IT infrastructure guidance, understand processes, troubleshoot issues and review architecture. You can also ask about AWS account details and your available SSO access roles, subject to your permissions.
It covers Cloud, Devtools, IT Security and Integration Platforms, together with general IT infrastructure and services, and IU Teams and groups.
Can the agent make changes to our systems, or does it only provide guidance?
It provides guidance and performs supported information lookups. It cannot deploy resources, change configurations or grant access to your systems. Suggested changes must be reviewed and carried out through your normal approved processes.
Cloud Architecture reviews assess the material you provide; they are not scans of your live environment.
How do I ask a question to get the best answer?
Describe what you are trying to achieve, name the relevant service or environment, and include useful context or error messages without sharing secrets. Say whether you want a short explanation, a checklist or step-by-step guidance.
For example: “How do I request access to GitLab for a new team member?” You can add details or ask follow-up questions if the first answer does not address your situation.
What should I do if an answer is incorrect or the agent cannot help?
For an incorrect answer or missing guidance, use the thumbs-down button below the response and briefly explain what was wrong or missing. You can also clarify your question or add relevant context.
Who can I contact about the Infra Agent?
For general questions about the Infra Agent, contact Cloud Adoption via email at cloud@scania.com.
Data & Privacy
We retain certain information to operate the Infra Agent, understand how it is used, and improve its usefulness, accuracy and reliability. Here you can learn how we use and handle that information.
What information can I share with the Infra Agent?
You may share information classified as Internal or lower, in accordance with Scania’s information-classification rules. Information classified as Confidential or Secret must not be entered in questions or uploaded as files.
Never share passwords, access tokens or private keys, and avoid including unnecessary personal information.
If you are unsure whether information is suitable to share, contact Cloud Adoption before submitting it.
Why do we retain information?
We use questions, feedback and usage information to understand what people need, identify missing or unclear guidance, investigate problems, and improve the service. Technical information also helps us monitor reliability, response times and resource consumption.
Who can access retained information?
Access to retained information is restricted to authorized members of the Cloud Adoption team, which develops and maintains the Infra Agent. The team uses this information to support users, investigate problems and improve the service.
How does my feedback improve the service?
Feedback helps us identify incorrect answers, unsuitable sources, missing information and usability problems. Reviewing the associated question gives us context to understand what went wrong and prioritize improvements.
Submitting feedback does not automatically change the agent’s behavior or guarantee that an answer has been corrected.
Is my information used to train the AI model?
Improving the Infra Agent is different from training the underlying AI model. We use service information to improve guidance, information retrieval and application behavior.
The agent uses Amazon Bedrock. Inputs and outputs from Bedrock are not used by AWS or third-party model providers to train their models. AWS data-use explanation (opens in a new tab).
What happens to files I upload?
Review files are temporarily stored in private, encrypted storage so the agent can process your request. Attachment references expire after 24 hours, and uploaded files become eligible for automatic deletion after one day.
Review uploads are not automatically added to the agent’s knowledge base. Information described in your question or the generated answer may remain in the saved conversation or related records.
Does deleting a chat delete all related information?
No. Deleting a chat removes it from your saved conversation history, but it does not automatically delete related analytics, feedback, logs or backup copies. Those records follow their own retention and deletion processes.