An agent PREPARES THE REPLY.
An example workflow: the agent responds to an email. Your team decides on the reply.Incoming email
How is our website draft coming along?
From: Mr. Meier · Sample clientHi everyone, is there an update on the design of our website yet? I'd like to go over it with my team.Best regards Michael MeierThe agent at work
Understand the request.
The incoming message starts the workflow. The agent recognizes the client and the question about the current draft.Orbit · Read project data
Find the right project.
The agent reads Mr. Meier's website project, which it has been given access to: the new draft is ready for review.Axiom · Read documentation
Check the agreed details.
The project documentation adds the requirements that were discussed and a link to the current draft.Prepare a reply.
With this context, the agent writes a draft email. It's then passed to your team for review.Draft reply · Reviewed by your team
- To
- Michael Meier <m.meier@beispiel-kunde.de>
- Subject
- Re: How is our website draft coming along?
Hello Mr. Meier, The new website draft is ready. It incorporates the changes we discussed. We'd be happy to go over the current version with you and your team. Best regards Your project team
Not every workflow NEEDS AI.
When the rules are clear, a classic automated workflow can be the best solution: transferring data, processing media or triggering a defined next step. A workflow like this is transparent and easy to check.AI comes in when language, content or changing context need to be understood. Together, we distinguish between fixed rules, preparatory suggestions and explicitly permitted actions.Clear rules
A defined trigger leads to a defined step.Understand content
AI helps with classifying, writing or translating.Act with purpose
An agent uses approved tools in the existing system.AI that knows the context and sticks to WHAT'S APPROVED.
A chat on its own doesn't get any work done in your software. Through the Model Context Protocol (MCP), an AI agent gets defined access to the tools of your Aurolynx applications. It can only discover and use what has been approved for it.Your AI agent
Its own identityTargeted approvals
Tools · Data scopes · DurationAurolynx
Business rules stay in forceContribute in Orbit.
An approved agent can read projects and work packages, create new work packages, update existing ones and write comments. That way, a task you've discussed becomes an actionable part of your project.Explore OrbitMaintain knowledge in Axiom.
Read documents, gather context and suggest changes: an agent can support knowledge work in Axiom. Whether a suggestion may be applied or something may be published depends on the rights and rules of the space.Explore AxiomUse existing context.
Approved customer information and file metadata can make a task easier to understand. Access is limited to permitted actions and data – without giving an agent blanket access to all application rights.The connected systemYou decide WHAT'S POSSIBLE.
The central policy, the agent identity, the access key and the current permissions together limit access. An approval does not replace the business checks in the respective plugin.Access can be time-limited and revoked. Particularly sensitive actions such as publishing, sending or deleting must be selected individually. An approval is a permission – not automatically a human confirmation before every single call.01
Specific tools
Choose read, write and admin access precisely.02
Limited context
Only accessible data and approved instructions.03
Traceable actions
Activity logs record which access performed which action.04