Agents in business processes.
Example: a quote. A chat writes the text. An agent does the groundwork in your systems: reading the enquiry, finding documents, costing.
- one process in operation
- Agent with permissions and tools
- Audit trail for every run
Data and knowledge systems.
Product and master data often exist several times, with differing values. We store them so that an agent finds an answer with its source.
- ERPLead time: 3 weeks
- PLMLead time: 4 weeks
- File share“about a month”
- Email“can be faster”
- Lead time
- 3 weeks
- Source
- ERP
- As of
- 01.10.2026
“3 weeks, according to the ERP, as of 01.10.2026.”
- Data model
- cleaned master data
- access with source reference
Secure AI environments.
Hosting in the EU or on your premises, access rights per person and agent, approved models only.
- Environment in the EU or in house
- Permission concept
- List of approved models
Agentic Workbench.
The software for it: a dedicated AI agent for every person, isolated, with gateway and audit trail.
- PurchasingOwn agentCompare supplier quotes
- SalesOwn agentPrepare a quote
- QualityOwn agentWrite an inspection report
- Workspace per person
- Gateway
- Audit trail
- Runs in your cloud
Four stages.
From assistance for individuals to a process that runs through agents from intake to filing.
- 01AssistanceIndividual people, individual tasks, without system access.
- 02Agent in a processAn agent handles one workflow in one system.
- 03WorkbenchAgents for everyone, with rules, permissions and an audit trail.
- 04End-to-end processA process runs through agents from intake to filing. People decide at the start and at the end.
Talks in this field.
As a talk, keynote or training session, for leadership teams, teams and events.
What comes after ChatGPT: AI agents in mid-sized companies
Management and IT leadershipWhat changes from a chat window to AI systems that act, and what your own IT needs to prepare.
From prompt to process
Departments and teamsAutomating recurring office work: three processes as examples, and how to tell where it pays off.
AI that knows your business
Management and IT leadershipAfter the first automations: a memory for terms, customers and decisions, and rules that make every step traceable.
Questions and answers.
01What is agentic automation?
AI agents take over steps in ongoing processes: reading the enquiry, finding documents, costing, filing a draft. Unlike hard-coded automation, they cope with changing documents.
02Which systems can be connected?
Anything with an interface: ERP, PLM, document storage, email, databases and specialist applications. The Model Context Protocol (MCP) gives an agent exactly the tools it needs, and no others.
03What happens when an agent makes a mistake?
It can only act within its permissions, and at defined points a person decides before a result takes effect. The audit trail shows which step went wrong.
Which process costs you the most time?
A 30-minute call, online, no obligation. Together we check whether an agent can take it on.