AETERION

    Platform

    Platform features

    BRAIAN by ÆTERION ships its capabilities in eight families: process execution, agent architecture, policy and governance, audit and observability, knowledge and conversation, integration, environments and security. Every capability below states what it does and whether it is available or in development, with a link to the page that documents it in depth.

    Process execution

    Read more

    How a business process is defined, run and completed inside the runtime.

    • Steps, owners and conditions are configuration, readable by the people who own the process.

    • A running process keeps the version it started on; new runs use the new version.

    • The approver sees the step, its inputs, the proposed action and the governing policy, and can approve, amend or reject.

    • Retries and escalation

      Available

      The runtime retries a failed step and escalates to a person when confidence or policy requires it.

    Agent architecture

    Read more

    How agents are organised, authorised and kept attributable.

    • Domain agents

      Available

      Specialised agents cover eight business domains, from finance and legal to supply chain and procurement.

    • An agent inherits only the permissions of the step it runs, for the duration of that step.

    • Model, prompt and policy versions are pinned and recorded, so every behaviour change is attributable.

    • Model usage runs through the platform gateway or through your own provider contract, billed at cost.

    Policy and governance

    Read more

    How rules are written, versioned and enforced at runtime.

    • Policies carry authorship and history; you can diff two versions, promote one to production and roll back a change without touching process code.

    • Policies are scoped, so business units, regions or subsidiaries hold different limits under one platform.

    • A change reaches production only when the test suite passes the thresholds you set, compared version against version.

    • A wrong output in production is promoted into the test suite with one action, together with its inputs.

    Audit and observability

    Read more

    What is recorded for every run, and who can read it.

    • Inputs, outputs, cost, model version and approvals of every run are written to an immutable trace.

    • Cost is attributed per process, per team and per model, so budget reviews use measured numbers.

    • Trace retention is a policy setting, aligned to your records-management rules.

    • Traces export to your SIEM or data warehouse for long-term analysis.

    • Who may read traces, and which fields they see, is decided by the control plane.

    Knowledge and conversation

    Read more

    How people query running processes and organisational knowledge through Carol.

    • Ask a running process a question by voice or text, under the same permissions you hold in the console.

    • Answers are retrieved under the asker's permissions, so a question never reveals data the person cannot already see.

    • Carol answers in English, Italian and German, including mixed-language questions about the same process.

    • Carol works in Microsoft Teams and Slack through the existing connectors, not only in the console.

    • A conversation keeps the process it is about, so follow-up questions do not need to repeat identifiers.

    Integration

    Read more

    How the platform reads and writes in your existing systems.

    • Connectors for 16 enterprise systems cover ERP, CRM, collaboration and document platforms.

    • Custom connectors are built under the Scale plan for systems outside the catalogue.

    • Connector credentials are held in the control plane and never exposed to agent prompts or process definitions.

    • Additional connectors

      In development

      [DA COMPILARE]

    Environments

    Read more

    How development, staging and production stay separated, and where the platform runs.

    • Development, staging and production hold separate credentials and separate limits.

    • Moving a policy or process version from staging to production is an explicit, recorded step.

    • Managed hosting runs in Switzerland or the European Union, with data-residency options on enterprise plans.

    • Enterprise plans deploy the platform into your own tenancy, under your operating responsibilities.

    Security

    Read more

    Controls against the risks specific to AI systems in production.

    • Untrusted content is isolated from instruction context, and tool access is authorised by policy rather than by model output.

    • Redaction rules run before any request leaves the environment.

    • Model routing can be restricted by jurisdiction, so data stays within the legal regions you allow.

    • Users and service accounts authenticate through your identity provider; roles map to policy, not to individual scripts.