Message
An employee sends a description to an existing case through a private Teams chat.
My own Power Platform project
Service request preparation assistant
From a message linked to an existing case, the assistant proposes service, address, device and problem. Every proposal includes the source quote it is based on.
Copilot StudioAI BuilderPower AutomateSharePointTeamsApprovals
Request preparation assistant
Starting point
AI proposes. A person checks and approves.
Information needed to handle a case is scattered through message text. Retyping requires attention, and AI proposals should not be accepted without review.
The workflow
How it works
One problem, clear steps and a visible result.
Proposals with sources
AI proposes four fields. Rules outside the AI validate them before proposals are saved.
Human review
The case owner sees the values and quotes and personally makes the decision.
Controlled write
Only approved fields reach the record. Final readback confirms the outcome.
What the working solution shows
I verified the complete flow in my own environment using synthetic data: message, proposals, approval card and final readback. The demonstration shows human control over the information written to the case.
View demo PDF (Polish)What connects this project
Tools and their roles
- Copilot Studio
- The employee conversation and calls to the defined workflow tools.
- AI Builder
- Proposals for four fields based on the message, with supporting quotations.
- Power Automate
- Case retrieval, rule checks, review and the controlled update.
- SharePoint
- The case record, proposed values and workflow events.
- Microsoft Teams
- The employee’s private conversation with the agent.
- Approvals
- A person’s explicit decision before approved values are written.
The demonstrated and tested workflow runs in Polish, using sample data. It works with an existing case.
Fit
When a similar process may help
- The team retypes information from request descriptions.
- Every proposal must point to its source text.
- Writing data requires a decision by the case owner.
Scope and requirements
What we agree before deployment
The demonstrator operates in Polish and uses an existing case. It does not send customer messages. Before deployment we agree on the process, record permissions, licensing, AI costs and acceptance scenarios. AI can make mistakes; outputs need review.
Let's discuss your task
Have a specific task
or a process to improve?
Tell me how your team works today and what you want to change. You do not need a ready specification or technical knowledge.
We briefly check the fit. Detailed investigation, building or repair is agreed as paid work.
Write to Maciejkontakt@klimekautomation.plPrefer a form?
Use a general description without confidential customer data. Privacy policy.