Drop a file in, get insight out

The big unlock

📖 In Your first conversations you typed prompts. That is chat. The actual win happens when you give the CLI a real file to work with.

A 30-page customer RFP. A transcript from a discovery call. A pitch deck. A folder of partner solution briefs. The CLI can read all of that and produce something useful.

How "context" works

When you run copilot from a folder, that folder becomes its working directory. The CLI can read files in that folder by filename. You don't paste content. You reference it.

Example:

PS C:\Users\you\downloads> copilot
✦ Copilot CLI · ready
> Read customer-rfp.pdf and summarize the customer's top 3 asks for an exec audience.

  Reading customer-rfp.pdf (28 pages)...

  Top 3 asks:
  • Migrate 200 SQL Server instances to Azure SQL by Q3
  • Establish a 24/7 managed service tier for the production estate
  • Co-fund a $50K POC against the Fabric platform

> _

The CLI sees customer-rfp.pdf is in its working directory, reads it, summarizes.

⚠️ Forgetting which folder you're in is the number-one source of "it can't see my file". Type pwd (or look at your prompt) before you start. If your file is on your Desktop and you opened copilot from C:\Users\you, the CLI can't find it.

🚀 Hands-on: three exercises

Exercise 1: summarize a real file

📖 Pick a file you'd analyze this week. Anything. A customer deck. An analyst report. A long email thread you exported. A meeting transcript. Put it in a folder.

cd <folder containing the file>
copilot

Inside copilot:

> What's in <filename>? Give me a 5-bullet summary.

Tweak:

> Now identify three things the customer cares about that they didn't say out loud. What's the subtext?

This is where the CLI earns its keep. Pattern-matching across the whole document. Faster than you reading it.

Exercise 2: structured extraction

📖 Drop in a customer's website "About" page (save the HTML, or paste the URL, depending on your setup).

Ask:

> From this content, extract:
> - Company size (employees, geography)
> - Practice areas
> - Customer logos mentioned
> - Microsoft technologies they reference
> - 3 things they're clearly proud of
> Format as a Markdown table.

That is a pre-call brief, generated in 30 seconds. You used to do this by hand for every meeting.

Exercise 3: comparing files

📖 Drop two customer RFPs (or two customer briefs) in the same folder.

> Compare customer-a-rfp.pdf and customer-b-rfp.pdf. What are 3 themes they have in common? What's unique to each?

If you don't have two RFPs, use any two long documents.

⚠️ Data boundary reminder

When you drop files in:

• The file content is sent to the Copilot endpoint as part of the conversation.
• Treat this the same way you'd treat M365 Copilot.
• Public, internal-only, customer-NDA → usually fine for analysis you keep internal.
• Customer personal data, credentials, anything labelled Highly Confidential → don't.

If in doubt, redact first.

> What does this folder contain? Categorize the files and tell me what kind of project this is.

🎯 Real-work pick-up

📖 Pick the next customer meeting on your calendar. Find one document related to that customer (their website, a past email, an old deck). Drop it in a folder, ask Copilot for a 4-bullet pre-call brief. See if it saves you the prep time.

That is the test that matters.

✅ Ready to move on if

• You've successfully fed at least one real file to Copilot and gotten useful output
• You know how to set the working directory before starting
• You're comfortable asking for a specific output format