The AI makes the draft
It reads the documents, finds the source and prepares an answer, a letter or a proposal. Every statement names its source.
Workflows with approval
The AI prepares. A person decides.
Many processes in a business follow the same pattern: something comes in, someone checks, someone approves, something goes out. We build such processes so that the AI does the preparation and nothing leaves without approval.
What it is
A workflow with approval connects people, documents and AI.
Two everyday examples: a receipt arrives, the AI reads it and prepares a proposal, a person checks and approves. A client asks a question, the AI searches the file and drafts the answer, the practice checks, edits and sends it.
A person checks and decides
You see the draft with its source. You approve, have it changed or discard it. Only after your decision does anything go out.
Nothing goes out without approval
This applies to letters, bookings and everything that leaves the house. If a cloud AI is involved, only the checked shadow copy goes out.
How a workflow comes about
Five steps, one process at a time.
01
Pick a process
We pick a process that occurs often and costs time. One process, not all at once.
02
Define roles
Who checks, who approves, who is only informed. You decide, not the software.
03
Connect sources read-only
Mailbox, filing, meters, merchandise system: we read along, nothing is written. Where an interface is missing, we say so.
04
Trial run with real documents
For a week the process runs with your documents. You check the drafts; during the trial nothing is sent.
05
Operation and refinement
In daily use we look together at where drafts are often changed and improve them.
Traceable
Who approved what, and what went out?
A workflow comes with a trail you can read later.
The data recorder
It records what leaves the house, in a chain that cannot be changed afterwards without being noticed. It checks nothing, it only records.
To the data recorderRecord approvals
Who approved which draft and when, we record with you in the process. Which details exactly belong to it, we define in conversation.
The data recorder is connected in the test environment; connection to all routes is in progress. It is not a legal certificate.
Briefly answered
The questions that come first.
How is this different from classic automation (RPA)?
Classic software robots click through input forms. We read documents, prepare drafts with sources and let a person decide. If a task is better solved with a fixed program, we will tell you honestly.
Who approves?
The people you define. Approval can lie with one person or a group, depending on how your business works.
What if the AI is wrong?
Every draft names its source, so the passage can be checked in the original. A second system continuously looks at answers and reports anomalies; it changes nothing.
Does our data stay in the house?
In local operation, yes. If you also want to use a cloud AI, only a checked shadow copy goes out; that is what the Pseudomizer does.
Which systems can be connected?
Much can be connected read-only: meters, scales and controllers via Modbus or similar protocols, plus mailbox, merchandise system, accounting and filing. Which interfaces exist, we clarify in the inventory. Where there is none, we say so.
How do we start?
With one process and one source. If it runs for a week and holds up, the next one follows.
More
Related pages.
Connect data
What gets connected and how we proceed.
To the pageSecurity and documents
Check chain, data recorder and evidence for your data protection team.
To the pageCorporate apps
What runs the same every week does not have to be done every week.
To the pageBring us your biggest annoyance.
One conversation, one real task from your daily work. We use it to show what AI and automation can do for you.