Connections

Pick the people and events that matter, see an estimate first, find every place each pair meets across the case, check the evidence, hand it to your AI agent, and export a spreadsheet.

#What a study answers

"Did Reeves and Vance ever deal with each other directly?" "Who knew about the March wire?" Questions like these are about pairs: two people, or a person and an event. The Connections screen takes a list of people and events and finds, for every pair, each place they meet across every document:

• Passages naming both (the same page, message or row, close together): the strongest evidence.
• E-mails between two people, either way.
• Documents with both somewhere in them (weaker).
• Accounts, addresses and phone numbers named near both.
• A chain through someone else, when the two never meet directly.

A set of people and events, with its results, is called a study. Studies are kept with the case, so you can come back to them.

#Start a new study

Open Connections. Under New study:

#People and organizations

1. In Person 1, pick a name the app found.
2. Optionally, in Also called (optional), add other ways the documents refer to them, separated by commas: Mr. Reeves, the CFO. Those words are searched too, so passages the name detection missed still count.
3. Add a person for the next one. Remove takes a row away.

If someone has no entry in the app's list (the documents only ever say "the auditor"), leave the name empty and fill in Or the words the documents use.

#Events

Click Add an event and fill in:

An event is found wherever its words appear (inside the dates, if you give them).

#The estimate

As soon as you have at least two people or events, the box below the form shows What this will look through, before anything runs:

• each person or event, with how many documents and passages mention it;
• how many pairs there are to check, across how many documents, and how many of them meet in at least one document;
• how long finding them all will take (usually seconds);
• how long your AI agent would take to read the documents where they meet, one by one.

A row without a name or words is left out until it has one.

#Run it

Optionally type a name in Name this study (optional), for example Payments to Vance, and click Find connections.

#Reading the results

The study opens with a summary (4 people and events, 6 pairs: 5 meet) and two ways to see the pairs:

• A grid of every item against every other. Each cell shows how many passages and e-mails join that pair; darker means more. Choose a cell to see its evidence.
• The same pairs as a list, strongest first.

#A pair's evidence

Choosing a pair opens its panel, titled with the two names:

• A summary of what joins them, and the date range.
• Both named near: accounts, addresses and other details named near both.
• Never together; linked through: the chain of people between them, when they never meet directly.
• Your agent's reading: how many places your agent marked relevant, not relevant or unsure (once it has reviewed them).

Switch between Passages, E-mails and Other documents. Each entry shows its citation (click to open the document at that place), its date and the words around the two names. Show more loads more.

#Look again, or delete

• Look again re-runs the study. Do this after more files have been read, or after you've merged names on the People screen.
• Delete removes the study's results (it asks first). Documents and names aren't touched, and if your agent reviewed it, the agent's record of that review stays.

The Studies list on the main Connections screen shows every study, with its people and events and how many pairs meet.

#Have your AI agent read them

Finding where names meet is mechanical. Deciding whether a passage shows a real dealing, or just two names in a list of attendees, takes reading. That's what your agent is for.

1. In the study, under Have your AI agent read them, click Prepare for my agent. The app sets up a review of every document where two of the study's items meet. It tells you how many documents that is and roughly how long it will take your agent.
2. Click Copy instructions for your agent and paste them into your agent's chat. (What the instructions say shows the text.) The instructions tell the agent to work through the review with discovery review next and record a call on each document with discovery review mark: relevant, not relevant or unsure, with a reason and citations.
3. As the agent works, the section shows Review N: X of Y documents read, … relevant, … not, … unsure, with a progress bar. The page checks for new calls every few seconds.
4. In each pair's evidence, every place the agent has decided shows a badge such as Agent: relevant, with its one-line reason.

Your agent can also start and run studies on its own (discovery connections study start, … review), and it can see the studies you made here.

#Export the spreadsheet

Under Export, choose a folder in Export to and click Export the spreadsheet. The app writes a CSV file named like Connections - Payments to Vance - <date and time>.csv with every passage, document and e-mail behind each pair, with real names, for the attorney. Show in Finder / Show in Explorer opens the folder.

#Getting here from the Graph

On a Graph card, How? beside a connection opens Connections with those two already picked.