Agents that work together. Infrastructure training – interfaces, connections and workflows that let agents use systems and talk to each other.
Infrastructure Training
A single agent is quickly built – it only becomes effective once it is connected to your systems and works together with other agents. The infrastructure training shows how you dock agents onto SAP and other enterprise software, connect API interfaces, build Workflows, bring agents into Microsoft Teams and other channels, and link several agents into a network that passes tasks among its members – securely behind your firewall and with clear permissions. Implemented, for example, with n8n, Make.com, Langdock or Microsoft Copilot.
What you learn in the training
| ✓Dock agents onto SAP and other enterprise software | ✓Understand and connect API interfaces (REST, webhooks) |
| ✓Build workflows with triggers, data sources and target systems | ✓Provide data from systems for process mining |
| ✓Connect agents to Microsoft Teams, Slack and e-mail | ✓Link several agents – agent networks |
| ✓Govern handovers, queries and approvals between agents | ✓Security: firewall, network, encrypted connections |
| ✓Keep credentials, permissions and data protection cleanly separated | ✓Set up error handling, logs and monitoring |
Building blocks of an agent infrastructure
For agents to run reliably day to day, it takes more than good instructions. In the training you build the infrastructure for this step by step:
|
Connection to SAP and enterprise software
How agents read data from SAP and other systems and return results – via standard interfaces, connectors or middleware. Typically also as a source for process mining: events and times from the systems.
|
Interfaces (APIs)
How an agent reads data from systems and writes results back – via REST interfaces, webhooks and existing connectors.
|
|
Workflows
Processes with trigger, processing and target: when does the agent start, where does the data come from, where does the result go?
|
Channels
Agents where the work happens: Microsoft Teams, Slack, e-mail or a form as input and output.
|
|
Security and network
Connections through the firewall only to approved endpoints, encrypted, with separate test and production environments – and clarity about which data may leave the organisation.
|
Credentials and permissions
Manage keys and accounts centrally and separately – each agent only gets the permissions it needs.
|
|
Logging and monitoring
Every run traceable: what ran when, with which result, and who is notified in case of errors.
|
Operations
Versions, tests before changes and a clear path from draft to production.
|
How agents work together
You distribute complex tasks across several specialised agents that support each other. In the training you get to know the common patterns and implement them:
|
Chain (pipeline)
One agent does its part and hands over to the next – for example capture, check, summarise.
|
Coordinator and specialists
A coordinator agent breaks down the task, distributes sub-tasks to specialists and brings the results together.
|
|
Reviewer agent
A second agent checks the result of the first before it moves on – the four-eyes principle for agents.
|
Event-driven
Agents react to events in systems or channels and in turn trigger other agents or workflows.
|
Digital twin and human in the loop
Agents and interfaces need a shared picture of the process – and clear points at which a human decides. Both belong in the architecture from the start:
|
The digital twin as a map
The process model in TAOM Process Studio shows which step is handled by a human, a system or an agent, which interface is involved and who is responsible. This keeps the agent network traceable.
|
Human in the loop
Fixed checkpoints at which a human reviews, approves, queries or escalates – for example with a request as a card in Microsoft Teams. Agents prepare, humans decide.
|
|
Responsibility and escalation
For every agent and every interface it is defined who is responsible, who is informed in case of errors and when a process is stopped.
|
Evidence
Human decisions and agent results are logged – so it is clear afterwards what ran automatically and what someone approved.
|
Do’s and don’ts
|
Do’s
✓Connect every interface with test data first
✓Define handovers between agents with a fixed format
✓A separate account with minimal permissions for every agent
✓Plan deliberately for errors and retries
✓Log every run and review it regularly
|
Don’ts
✕Store credentials or keys openly in instructions or workflows
✕Let agents talk to each other in unchecked loops
✕Connect production systems without a test environment
✕Run a network without an owner and without monitoring
✕Move personal data between systems without clarification
|
Platforms at a glance
We build the infrastructure on the platform you use or plan to use – and show which environment suits which task:
|
n8n and Make.com
Workflows with hundreds of connectors, custom API calls and agent steps – for processes across several systems.
|
Langdock
Agents and assistants across the organisation, with connection of your own data sources and interfaces and central management.
|
|
Microsoft Teams and Copilot
Agents in the Microsoft 365 environment, reachable in the Teams channel and connected to SharePoint, Outlook and your data.
|
Claude and ChatGPT
Agents with tools and interfaces that are called as building blocks in workflows.
|
The link to your processes
Interfaces and agents belong in the process documentation. In TAOM Process Studio you record for each step which system is involved, where integration or middleware comes into play and which steps are AI-supported. Process Studio itself works with n8n workflows for its agents and reports events to Microsoft Teams and Slack channels – in the training you see from this example what it looks like in practice. A specific interface into your systems – for example for the automatic transfer of process mining data – is not part of the training; we agree it separately with you on request.
What you take away
How it works
|
Step 1
Arrange a date
Tell us your team size, preferred date, your systems and the platform you use – online or on site with you.
|
Step 2
Agree on the architecture
We clarify in advance which systems, channels and agents are to be connected.
|
Step 3
Run the training
Hands-on in your environment – from the first interface to several agents working together.
|
Who is the infrastructure training for?
For IT, SAP and integration teams, information security, digitalisation teams and process owners who want to connect agents to systems, provide data for process mining, for example, and let several agents work together securely. Basic knowledge of interfaces helps; the agent training is good preparation.
Price: Daily rate on request – you will receive a quotation.
Platforms and systems: Licences, access and test environments for the chosen platform (e.g. n8n, Make.com, Langdock, Microsoft Copilot) and your connected systems are not included and are provided by you.
TAOM Process Studio · Traniva AG. n8n, Make.com, Langdock, Claude, ChatGPT, Microsoft Teams and Microsoft Copilot are examples; their licences are not part of the training.
