TAOM AI Agents™
Seven specialists instead of one all-rounder.
Instead of one general assistant for every task, TAOM uses specialised agents with defined responsibility – each at the level it can oversee, and all within the same company context.
- Specialised
- Shared context
- People decide
Business language is enough
Explain, don’t prompt
Anyone who knows a process describes it correctly on their own – in full sentences, not in instructions to a machine.
“A customer places an order. Sales checks availability. If the product is not in stock, the demand goes to production. Before shipping, quality assurance checks the finished product.”
From these four sentences the agent reads out structure:
- Activities
- Roles
- Decisions
- Systems
- Business objects
- Requirements
- Dependencies
Nobody has to learn BPMN first. Spoken language, workshop notes and work instructions are also valid inputs.
The seven agents
Nobody has to know everything
Each agent works at a specific level: on the diagram, on the single shape, on the phase, on the whole process – or on the person sitting in front of it.
Process Agent
Builds the model’s structure: roles, systems, transactions, inputs and outputs, risks, controls, key figures.
Application Agent
Establishes the application link – process step ↔ system ↔ specific transaction. Vendor-neutral as long as no provider has been explicitly approved.
Quality Agent
Per step: requirements, risks, controls, tests, evidence, approvals. Also watches for neutrality, so it does not rate providers. The tests themselves run through testing and verification.
Knowledge Agent
Finds and organises knowledge from documentation, decisions, specifications and experience – for example from Confluence or SharePoint.
Delivery Agent
Translates business needs into epics, user stories, tasks, functional specifications, test objects, dependencies and deliverables.
Collaboration Agent
Reads meetings, minutes and messages and transfers confirmed decisions, tasks, risks and open items into the target systems.
Learning Agent
Adapts explanations, guides and learning content to role, prior knowledge and work situation.
Orchestrated, not stacked
The Process Agent does not have to do the Delivery Agent’s work, and the Delivery Agent does not have to become an SAP expert. All contribute to the same context.
Assignment
Not everyone needs all of them
Which agents work depends on the role in the tool – set in the admin panel, not negotiated per person.
Context instead of prompt
Better environment, better answers
A general assistant starts at an input line. An agent in a company should know more before it starts. TAOM can give it:
- Process
- Process hierarchy
- Role
- System
- Requirement
- Business description
- Process phase
- Classification
- WRICEF object
- Governance information
- Knowledge reference
- Delivery information
This way an agent works within the context of the organisation instead of treating every request as a detached conversation.
Responsibility
The agent suggests, the human is accountable
Questions instead of inventions
If a description is ambiguous – for example several independent flows with no recognisable connection – the agent does not keep building on a hunch. It asks a question with concrete answer options.
Question and answer remain readable in the history. That way it is traceable later on what basis a model was created.
You decide how long to wait
The admin panel has a grace period. If the field is left empty, the editor waits indefinitely for the answer – the decision explicitly stays with people.
A number in it means: after this time the agent continues building. This is deliberately configurable and not the default.
From assistant to workforce
Six details turn a helper into a role
An assistant helps one person. An AI workforce carries the organisation’s work – for that every agent needs the same details a position would have.
Getting started
With one process and one agent
Describe how the flow works. Let it structure it. Look at the result. And only then add further agents, knowledge sources and context.
Distinction
What sets an agent apart from a chat window
“AI in the tool” today usually means: an input field that answers everything. A TAOM agent is something else – it has a position in the operation, not just a text field.
A copilot
Answers what you ask. Without a role, without responsibility, without limits – what it may do is decided by the prompt.
A TAOM agent
Has a Role, a task, a knowledge domain, Tools, powers – and human approvalwhere it matters.
That is why seven agents instead of one all-rounder: whoever may do everything cannot be controlled. And what an agent generated remains labelled on the element – readable, not merely claimed.
In the TAOM framework: MGT-08 – the methodology behind this function.
Where the agents get their context
The agents do not work in a vacuum: the stored models create a library of recurring roles, activities and areas – including the style in which this company models.