Advanced scenarios and agents
Advanced scenarios and agents
From assistant to agents
graph LR
C["Chat<br/>question → answer"] --> R["Researcher<br/>multi-step research<br/>work data + web"]
C --> A["Analyst<br/>Python calculations<br/>over data"]
C --> N["Notebooks<br/>project space<br/>with sources"]
C --> P["Pages<br/>shareable output<br/>to co-edit"]
C --> B["Your own agent<br/>Agent Builder<br/>e.g. 'SM-07 assistant'"]
C --> W["Cowork<br/>delegate a bigger task<br/>across apps"]
| Tool | How a manager uses it |
|---|---|
| Copilot Chat (Work) | overview across mail, meetings and files: “Prepare me for Thursday’s meeting.” |
| Researcher | decision support: bid/no-bid, supplier analysis, market |
| Analyst | calculations, forecasts, statistics over files (CSV, Excel) |
| Notebooks | one space for the D1 WIM project – files, e-mails, meetings, notes |
| Pages | chat output as a living document you share and extend with your team |
| Agent Builder | your own agent over company documents (policies, price lists, procedures) |
| Cowork | “do it for me” – several steps across apps with your approval at the end |
Copilot as decision support
- Collect – what we know (e-mails, documents, numbers)
- Structure – criteria, options, risks
- Challenge – “What speaks against it?”, “What would the CFO say?”
- Decide – you, not Copilot
Warning
An agent works longer and with more sources – which makes checking the citations even more important. With agents over company documents, think about who has access to them.