Documents & PDFs

One of the most immediately useful AI capabilities is simply that you can upload almost any document — a contract, a scanned manual, a slide deck, a dense report — and it will read it. This removes a huge barrier that used to exist: needing time, focus, and patience to get through dense material.

This works even for scans and photographs of paper documents, not just clean digital PDFs, thanks to the same vision capabilities covered in B1. That means the pile of paperwork in a drawer is just as accessible as a downloaded PDF.

The mental shift here is treating "I don't have time to read this" as solved by default, rather than as a reason to procrastinate on an important document indefinitely.

The most effective way to approach a long document isn't reading it start to finish — it's layering your understanding. First, ask for a short summary (100–200 words) to get the gist. Second, ask for the structure — what sections exist and what each covers — so you have a map. Third, ask targeted questions about the specific parts that matter to you.

This layered approach mirrors how experienced readers actually process long documents, but it removes the skill requirement — you don't need to be good at skimming, because the AI does that layer for you and you're just choosing where to drill in.

The order matters: jumping straight to targeted questions without the summary and structure first often means you don't know what to ask about yet.

For documents with a specific concern in mind — a lease, a contract, a policy — extraction is often more valuable than a general summary. Ask for every clause related to a specific topic (deposits, termination, liability) and you get a focused, comparable list instead of having to hunt through pages yourself.

This is a fundamentally different (and often faster) way to find information than reading — you're not looking for the needle in the haystack, you're asking the AI to pull out every needle at once.

Asking for quotes with clause or section numbers turns this from "trust me" into something verifiable — you can jump straight to the source location and confirm the extraction is accurate.

AI is genuinely useful for a first-pass contract review: flagging unusual terms, listing obligations, and generating a list of questions worth asking a landlord, client, or counterparty. This turns a 30-page contract from an intimidating wall of text into a short, actionable list.

But a first-pass review is not the same as legal advice. AI doesn't know your jurisdiction's specific enforceability rules, can't represent you, and can miss context a licensed professional would catch. Its real value is in surfacing questions and unusual clauses efficiently — it makes the eventual conversation with a lawyer (if the stakes justify one) much shorter and sharper.

The right mental model: AI gets you 80% of the way to knowing what to ask, and a human professional handles anything that's actually binding and high-stakes.

Contracts, policies, and agreements often go through revisions, and spotting exactly what changed between two versions by eye is tedious and error-prone. Upload both versions and ask for a comparison — ideally structured as a table of clause, before, after, and why it might matter.

This surfaces subtle but consequential edits that are easy to miss manually — a single word changed in a liability clause, a deadline quietly shifted — because AI compares systematically rather than relying on you noticing a small change buried in a big document.

This skill generalizes beyond contracts: comparing draft versions of any document (a policy, a proposal, a report) works the same way and is one of the highest-leverage document tasks for anyone who negotiates or reviews agreements regularly.

A summary isn't one-size-fits-all — the right summary depends entirely on who's reading it. A summary for your boss might focus on decisions and risk; a summary for a client might focus on what it means for them; a summary for your mom might need to skip jargon entirely.

Specifying the audience explicitly in your request ("summarize this for someone with no background in the topic" or "summarize this the way you'd brief a busy executive") produces a dramatically more useful result than a generic summary.

This is a small habit with a big payoff: you can generate multiple audience-specific summaries from the same source document in minutes, something that would otherwise mean writing several separate briefs by hand.

After any summary or review, one question consistently surfaces things a straightforward read wouldn't: "what important thing does this document NOT address?" This reframes the AI's task from describing what's there to actively looking for absence, which is a different and often more valuable kind of analysis.

Gaps matter enormously in practice — a lease that doesn't address subletting, a contract silent on what happens if a deadline is missed, a policy that doesn't cover an edge case you're worried about. These are exactly the things a normal read-through tends to skip past, because there's nothing on the page drawing your attention to them.

Making this question a habit — asking it after every important document review — is one of the highest-value single moves in this entire module.