This one has been blowing me away…



It's called Fyxer, the AI software studies your writing style and patterns when responding to repetitive emails. When you receive an email, it automatically drafts a response within outlook or gives you options for the software schedule meetings on your behalf.

With time, the draft emails continually gets better. In my profession, I'm responding to 100 emails per day and already cut time needed in half. As each week goes by, the number of emails it drafts needing manual correction becomes less.

I am defending a case against a Mexican grocery store for a slip and fall. (She is on video looking at her phone when she tripped over a box on the floor. New items were being shelved but the employee walked away for a second.)

Anyway, I did insurance defense for years, so I can do these depositions in my sleep. I now mostly do Plaintiff cases, but this client had a glitch with insurance, so I've had to step in as a defense lawyer again. But I wanted to check what AI would do.

I input the complaint, her discovery responses, the video, her medical records, etc. into Google Gemini (paid version - deep thinking mode.) I asked it:

Quote:

---

I represent the defendant, a hispanic grocery store. The plaintiff claims to have slipped and fell on a pallet or stack items to be shelved. The video and screenshot show she was on her phone at the time. Also, she provided pictures of a huge cut on her leg that apparently she is trying to claim was caused by the fall, but her medical records do not. Draft a deposition outline for the defense attorney, covering her background, litigation history, recall of the accident, medical history and medical and bills relating to the alleged fall. Include detailed questions about the incident.

---

I have a standard depo outline I've used for decades in these types of cases. It lacked the detail of that, but it was exceptional at pulling out all the dates, names, etc.. - basically all of the facts that I would have to review the file and pull out. Saved me hours of prep time.

The deposition was taken via Zoom and I have 3 monitors. I kept the Gemini thread up on one monitor and during the deposition, I was able to input her answers and it would generate additional questions and ideas based upon its "memory" of all the documents.

I also have an A.I. app on my phone that transcribes and summarizes the conversation real-time. Truly a game changer. It's like having an experienced legal secretary or associate with you at the deposition.

Personally, I think the best bet with AI (for almost everyone) is focusing on getting the AI to do the "work before the work."

No matter what you're doing, there's a bunch of stuff that has to be done before you do your thing. So don't get the AI to "do your job." Have it do everything that needs to be done before that.

My answer has been to build out an AI team that sets me up.

INTAKE AGENT. Takes raw, unstructured inputs from me and get them to the right place in the system. For example, I dictate a long, sloppy voice memo mentioning that I need to get buy eggs at the store, set a meeting with my banker, and I have an idea for a new product we should consider. The intake agent breaks that up into chunks and sends each one to the right place in the system.

LIBRARIAN. Manages a personal wikipedia. Non-actionable items that might be useful later. Articles, notes from conferences, podcast summaries, social media posts.

PROJECT MANAGER. For anything that's a "project" -- meaning it isn't going to be done all at once, is likely to have wait states and dependencies -- the project manager writes a project file, tracks related action items, and maintains a dashboard of all open projects.

ACTION MANAGER. Similar to the project manager, but deals with to-dos. Maintains to-do lists and a dashboard of all open action items (including "waiting for" items for things that have been delegated).

SYSTEM ENGINEER. Kinda like the project manager but specific to any items that are about the system itself (bugs, enhancement requests, feature requests). Maintains a dashboard of potential system fixes and improvements.

CHIEF OF STAFF. Has visibility into everything in the system described above, plus access to calendar and email. Writes morning briefs and afternoon debriefs.


HOW THIS ALL ENDS UP WORKING TOGETHER:

- Each morning I get a daily brief from the chief of staff. Basically some news plus a look at what's coming up in my day. Appointments, to-dos coming due, projects that are coming due, delegated items I need to nudge.

- Based on what I get in the brief, I'll come up with what I call the "day plan." This is my opportunity to tell the system "I'm going to do this, not going to do that, make this change, block this time, etc," The system updates things based on what I tell it.

- I go about my day. I capture things as I go. I have a few ways to put things into the system but I usually will just message it with Telegram. These are usually little reminders like action items that came out of meetings, random things I think of, errands I need to run, whatever.

- Afternoon I get a debrief. It basically is a quick recap of the day plan I committed to in the morning plus a look ahead at the next few days.

- I let the system know what I actually did. "Ended up not going to this meeting, didn't get X done, Y got finished, Z got pushed to next week." The system updates based on what I tell it.

- Repeat.


So the system isn't really doing my job, but it's handling a lot of the administrative stuff that makes it a ton easier to do my job.