The Document She Had Never Seen
More than once at a deposition, I’ve handed a witness a paper document and asked a simple question: “Can you identify this document?”
She, my deponent, looks at it. “No. I’ve never seen this before.”
We both know it is hers. Sometimes she is hiding something. Sometimes she truly doesn’t remember or recognize it. Either way, the pages in front of her have been cut loose from the day they were written. The document has no hour, no cabinet or folder it came from, and no sense of why it was written or for whom or where it traveled. She isn’t looking at something from her life. She is looking at an object presented alone on a stage of its own, as though it had dropped in unannounced.
The Breadcrumbs in a Document Collection
This week my class worked on collection, the step where information leaves the custodian’s world and begins to enter the case. We usually describe the goal in terms of content: get the files, keep them intact, prove nothing changed.
There is a second way of looking at it. A properly collected electronic file carries more than the words —body text— that would appear if it were printed on paper. The electronic file may carry the date it was created, the hour it was last changed, the device that produced it, the folder it sat in, and who authored and received it. Each is a breadcrumb, a small trace of a day the custodian lived with the words of the document. Think of the document as part of a human totality, with connections that light up its meaning and the meaning of the world around it.
Open a photo on your phone and swipe up. You’ll see when it was taken, with what, and where. Who doesn’t look at a photo’s metadata when a memory won’t quite come back? The memory is not only the picture. It is the context of what happened, to whom, when, and why. ESI gives us little gifts, and little gifts can make all the difference.
Now take a screenshot of the same photo. The picture survives and the path back to the richness the photo discloses is gone.
That is how digital breadcrumbs get lost, often by someone trying to help. A conscientious employee, told to gather “the relevant files,” drags them onto a thumb drive. The drive is copied, then emailed, and the dates and trail quietly change along the way. This is why we say the custodian shouldn’t collect her own data. She knows where her work lives, and our interview last week is how we learn that. But once she handles the files, she has changed them, changing what they disclose about her and others. Careful collection protects the trail with tools like hash values, a digital fingerprint showing that the file that arrived is the file that left.
One Crumb Is Helpful, but It Is Not the Whole Neighborhood
A single timestamp means very little. It means something next to others: the calendar entry that morning, the chat message an hour earlier, the draft saved twice before the final, the people copied and the one who wasn’t. Together, they begin to form a path through someone’s week, a week of work on case issues, but also work with others. I don’t think I ever had a case with one player. We live inescapably in a world with others, where we do our work with others, against others, pushed by others, and thanked by others.
We have never been able to pull the multitude of breadcrumbs and text together easily. Search tools gave us fragments, hits without their surroundings, like a bag of crumbs with few or no ways to see which ones lay side by side. Yes, search improved over the years, providing relationship wheels and other helpful reconstruction tools. Hence the traditional reliance on depositions. But the human in a deposition is its own struggle of forgetting, dodging, weaving, and sometimes even lying.
Today, AI has changed things. You can ask it to build a timeline from the collected material. You can ask where two accounts disagree, what else was happening that week, and where the record goes quiet. And you can keep going, because it is a dialogue. It resembles taking a deposition of the data, with a witness who is tireless and cooperative, though one whose answers you still check against the documents.
I want to be clear about what the dialogue is for. It isn’t a way to have the machine tell us what our witness was thinking. It is a way to stretch our own imagination so that we take up her position, see her desk and her deadlines, and ask better questions. In my article this past June, “The Recovery of Meaning in a Digital Age: E-Discovery and the Renewal of Judgment,” 30 J. Tech. L. & Pol’y (2026), I argue that AI can help recover the context that judgment depends on. AI can illuminate, but it cannot judge. It can measure and calculate. It can recall and suggest. But it can’t tell me whether to accept the settlement, whether to call a dangerous witness, or whether a request for social media is fair and proportional. And before any client material goes into such a tool, make sure it is one your firm has approved for confidential and privileged data.
The Map and the House
Last week I described the org chart as a map and the custodian’s working life as the house. A timeline built from breadcrumbs is a much better map. It is still a map, made of traces of her days and not the days themselves. If I walk into the deposition convinced that I know what she thought, I have mistaken the map for the house.
The work isn’t finished when the crumbs are gathered. It continues in the room with her. The path I’ve pieced together gives me better questions. I can ask our witness, “This was sent at 11:42 on a Thursday, from your phone, after three edits. Walk me through that night. Who were you on the phone with?” And so it goes, but not blindly.
And she may tell me: the deadline, the call from her manager, the reason she was still up. At that point the page stops being an exhibit between us. It becomes a way back into something that happened. Breadcrumbs matter a lot. Even more important, breadcrumbs together matter most.
A Friendly Reminder
Protect the crumbs: ask for natives, and don’t drag, screenshot, or forward ESI. Use the tools to put the pieces together and use the AI dialogue to widen your view of her situation. Then treat what you have built as a set of questions and take them to the person who lived it.
