Automation in criminal defense:

• Drafting routine pleadings/motions

• E-filing routine pleadings/motions

• Auto-organizing Discovery

 

AI in justice system/criminal defense:

• Police robots (yes - call me to chat)

• Axon Draft One - drafts police reports from body camera footage

• Reviewing & tagging body camera footage; writing transcripts

• Drafting case-specific, complex pleadings/motions (one day, with enough data)

 

AI and automation are different in a few ways:

• Tasks

• Results

• Error Handling

 

Tasks

​

AI: creative tasks like writing code or blog posts, summarizing writing

Automation: mindless, repetitive tasks like data entry

 

AI is an effort to teach computers to think and create like humans. 

​

Automation is delegating tasks to computers they are already better at than humans, so humans are freed for work only they can do.

​

​AI gets a lot of buzz for what it can potentially do.

Automation already does exactly what computers were always meant to do: logic.

​

Automation is concrete, if/then logic. Examples: 

​

If client paid, then draft and file my pleadings.

If it is a criminal case, then draft and file my Appearance, Discovery, Speedy Trial, and Suppress.

If it is a DUI, then file the above and Breath Tech/Chemist.

​If it is a District case, then use Template A. If Circuit, then use Template B.

​

Automation code is the same instructions you would give a legal assistant, in the computer's language.

 

A better word for AI’s current capabilities is “natural language processing.” “Intelligence” implies critical thinking. It doesn’t do that.

​

Currently, AI can summarize a document served to you or write a transcript of body camera footage (natural language processing). Long term, after thousands of examples of a certain type of evidence, it may be able to find the most important detail in a body camera video or draft a complex, case specific motion.

​

Results

​

AI is probabilistic. Automation is deterministic. 

 

Automation will do the same thing every single time. AI will not. 

​

Your e-filing bot is going to draft your Appearance & Discovery the same way every single time and file them into all related cases every single time. 24/7/365. No surprises.

 

Industry-relevant AI example of probabilistic software: Axon Draft One

​

At their 2024 conference in Miami, Axon CEO Rick Smith announced Draft One, an AI tool to draft police reports based on body camera footage.

 

It took a few times to phrase the question in a way that was clear to Axon’s software engineers, but finally the prosecutors asked the engineers, “When given the same video, will the AI produce the same report every time?”

 

The engineers said “close.” The prosecutors collectively groaned and said it was inadmissible. 

 

Furthermore, the AI is making the call as to whether someone “left” the scene or “fled” the scene. The officer SHOULD be reviewing and editing the report, but does not have an incentive to do so.

 

Error Handling

​

Neither AI nor automation can perform tasks that require human decision making.

 

Automation does not take on those tasks in the first place. It only performs tasks that can be reduced to if/then logic. A human designs the logic tree one time, and there are no new decisions to be made unless you refine or expand your bot.

 

Automation errors are lower quantity, higher quality errors than human errors.

 

For example, with our e-filing automation, 99% of the time, there are no errors at all. Our goal is to hear from a client for customer service no more than once per quarter. Low quantity of errors. 

 

If there is an error, the bot pauses, emails the CEO who makes a live fix (5-15 min), and the bot proceeds. It’s like a legal assistant who knows they are about to make a mistake, pauses, asks for help, then proceeds without error. The bot will not misspell a client’s name or file into a case you didn’t ask it to. High quality errors.

 

Automation can’t do everything, but what it can do, it does perfectly and inexpensively.

 

Rather than intervention by the CEO for benign errors that occur once in every few hundred citations filed, AI requires human review of each task it performs to eliminate errors.

​

For example, if we build AI to review Discovery for you, you still have to review it, but the AI may save you time by flagging events you listed during the client's intake.

 

Potential

​​​

I just wrote, "automation can't do everything." It can do a lot, though.

​

​~50% of administrative work can be automated (McKinsey & Co.). 

​

Automation is underutilized. 

​

Why? 

​

The people who can automate (software engineers) have never seen admin work and the people who do admin work (everyone else) do not know what can be automated.

​

Add in the risk and discomfort of starting and maintaining a business and Auto Admin's competitive advantages of customization, competence, ethics, persistence, and communication, there's not too many little Beccas running around conferences offering to build custom robots instead of hiring staff.

​

There will always be more automation work because new software is perpetually being developed and adopted and automation allows programs to speak to each other.

​

E-Filing: Case Search talks to Dropbox and MDEC.

 

Organizing Discovery: Axon talks to Email talks to Dropbox.

 

Build new AI software to review Discovery and forward it to a client? Needs to talk to Axon, Dropbox, and Email. More automation bots to build.

 

Auto Admin builds the "talks to." You build the rules for how to talk.

 

Police Robots

​

The police are understaffed and looking to build robots and optimize their budget to be more efficient just like everyone else.

 

Axon CEO Rick Smith is designing drones and police robots with the LAPD.

 

The Deputy Chiefs of LAPD told me excitedly, “With drones, we can arrest people in 3 minutes instead of 3 weeks!! You can’t shake a drone!”

 

The annual Axon conference is my #1 recommendation to stay ahead of technological criminal defense changes, given that police tech runs downstream to you.

 

Police & prosecution are dominating tech development and there is an enormous opportunity and ethical obligation for criminal defense to keep up.

 

Don’t be scared!

​

If tech can send retainer agreements, collect payment, draft, file, organize Discovery, and review Discovery, what will I do?

 

Sell. Cash checks. Show up in court.

 

Make people feel safe. Robots will never be able to do that.

 

Robots can’t build relationships with prosecutors and judges either.

 

You also do a lot more than I know about because I am not a lawyer. 

 

Furthermore, tech strategy and vision is easy to see and design. It may take a decade to execute.

 

If you’re within 10 years of retirement, these tech changes will increase your team's time, money, morale, and retention and decrease their errors, but you may not see police robots before you retire. They have liability risk and bureaucratic approvals to overcome.

 

You do need to be aware of tech like Draft One so you can defend against it. Depends how quickly it is adopted. It's currently being piloted in California, Indiana, and Colorado and the officers love it.

 

If you’re in your 30s or 40s, you will enjoy the above benefits, your team will (and should) end up being at least half robots and half humans and ideally, you are leading these market changes instead of being affected by them.

 

You cannot contain innovation. Drive it.