Every lawyer I’ve ever talked to has the same problem.
They get staffed on a new case or a new deal, and someone hands them a stack of documents, contracts, memos, correspondence, expert reports, and says “know this by Wednesday.”
So they do what every lawyer does. They read everything. They take notes. They highlight. They cross-reference. They try to hold the whole picture in their head at once. And somewhere around hour six, they’re buried in tabs, sticky notes, and caffeine, trying to remember which document mentioned that one date that connects to that other clause in that other agreement.
It’s not just lawyers, either. Anyone who works with documents, consultants, analysts, project managers, compliance officers, has some version of this problem. You have multiple sources. You need to understand them together, not separately. And the synthesis takes forever.
I recently tested a Microsoft 365 feature that tackles this head-on, and the results were impressive enough that I wanted to share them. It’s called Copilot Notebooks.
When you use regular Copilot Chat in Microsoft 365, it pulls from your entire tenant: emails, files, Teams chats, the whole universe of your organizational data. That’s powerful, but it’s noisy. It’s like asking a question in a stadium full of people. You might get a useful answer, but there’s a lot of background chatter competing for attention.
Copilot Notebooks flips that model. Instead of searching everything, you curate exactly what Copilot should work with. You create a focused workspace, drag in specific files (Word docs, Excel spreadsheets, PDFs, PowerPoints, meeting notes, even OneNote pages) and when you ask Copilot a question, it only uses what’s in that notebook.
No random emails from 2019 bleeding into your results. No hallucinations from irrelevant data sitting in a forgotten SharePoint site. Just grounded answers from the documents you chose.
Think of it as a project brain. You give Copilot exactly the context it needs for a specific task, and it works within those boundaries. Everything it tells you traces back to a document you put in the notebook. That grounding is what makes Notebooks fundamentally different from a general-purpose AI chat.
The sweet spot? When you have a collection of related documents and you need to understand them as a whole, not as individual files, but as a connected picture.
To put Notebooks through its paces, I set up a realistic scenario: a fictional trade secret lawsuit where a small software startup is suing a larger competitor and their former CTO for stealing proprietary technology. It’s a classic David-versus-Goliath IP dispute, and it’s the kind of case where a new associate gets handed a pile of documents and told to get up to speed fast.
I created six documents that represent what you’d actually find in a case file like this:
Client intake memo with the allegations, witness list, preliminary legal assessment, and estimated damages.
Employment agreement the former CTO signed, including non-compete, NDA, confidentiality obligations, and invention assignment clauses.
IT forensic audit report showing 3.4 GB of data was transferred to a personal Dropbox account in the weeks before the CTO resigned, plus details on a secure deletion tool he ran the day before quitting.
Technical comparison analysis with a feature-by-feature breakdown of the startup’s product versus the competitor’s suspiciously similar new product, including a development timeline analysis.
Email chain with seven key communications telling the story chronologically, from the resignation through the discovery of the data theft to the moment the competitor started pitching the startup’s own customers.
Competitor’s marketing materials including product announcements, a technical whitepaper, and a customer migration guide containing the startup’s proprietary API endpoints and customer-specific configurations.
I added all six to a Copilot Notebook. It took about 30 seconds. Then I started prompting.
“ “I’m a new associate onboarding to this case. Based on all the documents in this notebook, give me a comprehensive case summary. Include: a chronological timeline of key events, the strongest evidence for each claim, the key witnesses and their significance, and any cross-references between documents that strengthen the case.”
It built a timeline by pulling dates from across all six documents: the suspicious email from the competitor six days before the first data transfer, the transfer dates from the forensic audit, the resignation, the competitor hire, the product launch, the customer solicitation. No single document contains that full timeline. Copilot assembled it by reading all six simultaneously.
It identified the strongest evidence and explained why each piece matters, not just listing facts but connecting them. The forensic logs prove what was taken. The technical comparison proves the competitor couldn’t have built their product independently. The marketing materials prove they’re using proprietary information they shouldn’t have.
And it cross-referenced the employment agreement against the actual events, mapping specific contract clauses to specific violations with the supporting evidence for each.
That kind of synthesis takes hours manually. Copilot did it in about 20 seconds. EX:
“ “Based on the employment agreement and the forensic audit findings, list every specific contractual provision that was violated, the evidence supporting each violation, and which document the evidence comes from.”
This is the kind of task that normally eats an afternoon for a first-year associate. Copilot mapped every relevant clause (confidentiality, non-compete, non-solicitation, return of materials) to specific evidence from the forensic report, the technical analysis, and the competitor’s marketing materials. Each violation came with the date, the document source, and the supporting data point.
It’s not just summarization. It’s structured analysis across multiple documents, which is exactly what makes case onboarding so time-consuming when done manually.
Prompt 3: Deposition Prep
“ “Draft the top 10 questions I should prepare for the deposition of the former CTO, based on the evidence across all documents.”
This is where the output got genuinely impressive. The questions weren’t generic. They were grounded in the actual evidence, referencing specific dates from the forensic audit, specific files that were transferred, and specific details from the competitor’s marketing materials that could only have come from proprietary sources.
Would an experienced litigator refine these? Absolutely. But as a starting point that gets a new associate 80% of the way there? This saves hours of work.