Investigation in the Context of Artificial Intelligence

Here's a picture of where we are right now. Robots talking to Robots…meaning:

A claims adjuster asks AI to draft a denial letter. The letter goes to a lawyer, who asks AI to summarize it. The lawyer then asks AI to draft the response. That response goes back to the adjuster…who asks AI to read it and flag the important parts.

So, yep: "I was born after Judgement Day, into a broken world ruled by the machines. The worst were infiltration units that posed as humans. We called them Terminators."

Sounds a little absurd. But that's honestly where many of us sit as a legal community right now. Every side has access to AI: plaintiff firms, defense firms, insurance carriers, adjusters. And more and more of what goes back and forth in a case is drafted by AI and then read by AI on the other end.

So what's left that actually moves a case?

The evidence. It always has been. I wanted to give some thoughts about how crucial it is to gather evidence at the case inception.

I also wanted to give a shout-out to an investigator I absolutely love working with: Jay Paulino. He's been focused on training and executing his investigations for years, including how to capture and copy EDR data, in addition to his work with capturing evidence with high quality using drone photography. He's also bilingual and very organized in his approach to gathering evidence. Check him out at https://jppinvestigations.com/

AI can write the story. It can't create the facts.

AI is very good at taking information and organizing it, summarizing it, polishing it. What AI cannot do is go back in time and capture what happened at the scene. It can't pull data off a car's black box that's already been overwritten. It can't find the witness who was standing at the bus stop and saw the whole thing.

If you don't have the facts, the best-drafted letter in the world is just…words. And the other side's AI may read it that way too.

Cases still come down to the same questions: what evidence do you have, and how did you capture it? That work happens at the beginning of a case. And it's time-sensitive.

The early investigation: where cases are built

Some of what I look for early in a case:

EDR (black box) data. Many vehicles have an event data recorder that may capture speed, braking, throttle, and seatbelt use in the seconds before a crash. Vehicles get repaired, salvaged, or sold, and that data can be lost. Capturing it early can be the difference between “he said, she said” and “the data says.”

Photos and video of the scene. With a car crash, that means the intersection, skid marks, debris, sight lines, and signage. With a dangerous premises case, it means the condition itself: the crack, the spill, the missing handrail, the lighting. Conditions get fixed, cleaned up, or changed. Surveillance video may be recorded over. If it isn't captured, it may be gone.

Witnesses. Finding the people who saw or heard what happened, and talking with them while their memories are fresh. A neutral witness with no stake in the case can carry real weight with an adjuster, a mediator, or a jury.

That's a lot of fact-finding and evidence gathering up front. It's not glamorous work. But it's the foundation for everything that comes after.

(I've blogged before about thoughts on the scene of a collision or a fall, for anyone reading who's been hurt. That post is here https://www.ganciesq.com/post/involved-in-a-car-collision-or-a-fall-here-are-thoughts-for-actions-on-the-scene)

Circling back: using AI smartly and safely

So once all that information is gathered, where does AI fit?

I think AI can be a great tool for helping tell the story. It can help organize a timeline from hundreds of pages of records, line up witness statements against the physical evidence, and draft a demand that reads clearly.

But AI only works with what you give it. And what you give it depends on how fast and how well the evidence was gathered. That takes systems.

Have the system in place before the case comes in. Black box data can be lost. Video can be recorded over. Witnesses move, change phone numbers, or just forget. If the plan for gathering evidence gets built after the case is signed, valuable time may already be gone. The intake, the checklist, and the people to call should be ready to go on day one.

Know who you're deploying, and for what. Different evidence needs different people. An investigator to canvass for witnesses and surveillance cameras. A specialist to download EDR data the right way, so it holds up later. Someone to photograph and measure the scene before it changes. Knowing who to call, and having them ready to move quickly, can make all the difference.

Organize what comes back. Evidence gathered fast but filed in a mess isn't much better than evidence never gathered. Photos labeled by date and location. Witness statements tied to where each person was standing. EDR reports kept with the vehicle they came from. When it's all organized, it's ready to use, for AI and for everyone else on the case.

Lead with the facts. If the other side's AI may be summarizing your demand, make the important facts impossible to miss. Something like: “The EDR shows the defendant was traveling 58 mph in a 35 mph zone, with no braking before impact.” Lines like that survive a summary, whether a human or a machine is doing the reading.

Remember who ultimately decides. At some point a real person reads the file: an adjuster with authority, a defense attorney evaluating risk, a judge, a jury. AI may sort the information. People still react to a story.

And that's the goal: a story built on solid evidence, told clearly enough that the other side reacts. Not to how polished the letter sounds, but to what the evidence shows.

My final thought

AI is changing how we draft, read, and communicate in this field. I think that's mostly a good thing, if it's used carefully. But it hasn't changed what wins cases. In a world where everyone has the same AI tools, the side with the better evidence may be the side with the better case.