Stop Sending AI Workslop to Your Lawyer

It’s hard to tell, but I think OpenAI and its competitors have mostly had a positive effect on my business. While my organic search traffic has plummeted, a lot of new clients tell me ChatGPT referred them to me. And around 15% of my clients for the past couple of years have been generative AI startups. 

So I should have no real complaints as somebody selling shovels to prospectors during a gold rush, or, less charitably, selling oxygen to people who are inflating another tech bubble. 

But I do have one big complaint: I’m drowning in workslop. It has to stop. 

This Kind of Slop Is Called Workslop Now

Recently I saw a story in Morning Brew (one of my favorite newsletters) about a Harvard Business Review article that used the term “workslop” to describe the kind of AI-generated garbage that people are sending their colleagues.

Because large language models (LLMs) are still extremely bad at performing most types of white collar work, workslop is mostly unusable and has to be edited extensively or even replaced by the person who received it. 

This means the person who sent the workslop is essentially passing the job they were expected to do onto the recipient. In the end, the final product is probably worse than it would have been without the help of AI, with no time saved in aggregate.

Two different, related types of legal workslop have been making my life harder for the past year or so: prospective client workslop and current client workslop. While there are major differences between the two — the latter is worse — they’re both coming from the same generative AI trough.

Prospective Client Workslop

Starting in 2024, I started receiving emails from prospective clients containing long lists of bullet-pointed questions like these:

• •Do you offer flat-fee packages? 
• •Do any of your packages include responses to minor office actions?
• •Do your packages include searches and legal opinions on my chances of reaching registration?

Not only are these questions answered on my website, but they are answered right away, at the top of the page. You don’t even have to scroll down to my extensive FAQ to answer them. Imagine how frustrating it is for me to spend so much time presenting all this information clearly on my website, only for people to send me emails with ten bullet-pointed questions just as pointless as the examples above. 

While I have always gotten inquiries from people who somehow found my email address while skipping all of the information on my website, the situation is much worse now. In the past, these people would send something short like, “What’s the cost, how long does it take, and what documents do you need?” I had template answers I would paste in response, along with FAQ answers I could insert. 

Now this type of person, instead of asking three basic questions that are answered on my website, sends a list of eight to 12 questions that are somehow even more aggressively ignorant. And they use a lot more electricity and water to do it. 

To add some irony, I think I actually coined some of the terms of art these inquirers are asking about  — including ”major” and “minor office action” — when I first launched my firm in 2017, before they became standard marketing terms in the online flat-fee trademark attorney industry. I started using these terms because I thought words like “substantive,” “non-substantive,” and “procedural” were too long and confusing. 

These chatbots seem to be grabbing and appropriating those terms and a lot of other language about trademark services from my website and from competitors who adopted my language. So not only are people asking questions that are answered on my website, but they are grabbing terms I invented to describe my services and vomiting them back out at me to ask if I offer them. 

It’s insulting and obnoxious. Why should I spend 10 minutes writing answers to these questions you spent a few seconds generating when you didn’t even bother to spend a few more seconds looking at my website? Or worse, you did read the answers to these questions on my website and decided that you couldn’t be bothered to edit those questions out of slop you generated with a chatbot.

Current Client Workslop

I have no real obligation to reply to prospective clients who send me workslop, aside from my obligation to earn money to pay for food and housing. But when existing clients who have already paid for my services email me, I do need to respond. Even if they send me pages and pages of slop.

If you’re my current client and we’re communicating about your trademark, please don’t use generative AI to write your emails to me. And please don’t ask ChatGPT or another AI large language model about your legal problem and paste the answer into your email to me. These kinds of emails make it really hard for me to do my job.

Unlike most law firms, my firm doesn’t bill by the hour. My flat fee model keeps my pricing competitive and transparent while allowing me to work efficiently. I am generally happy to answer any extra legal questions you (my human client) have for me, but I will not answer a string of nonsensical bullet points that you generated in a few seconds using ChatGPT.

If I responded to everything my clients generated with AI, my business model would no longer work because I’d have to waste significantly more time on each project than I did before ChatGPT became popular. It’s not fair to my other clients for me to spend so much time to answer largely irrelevant questions that no human would ever think to ask me. 

Just as vibe coding is likely not resulting in better outcomes in the software industry, vibe emailing is not making you a better client.

The analysis offered by ChatGPT is almost never relevant to my clients’ specific situations, and often isn’t even relevant to the trademark application process at all. ChatGPT uses the web and other publications as its data sources, not real trademark application filing data. It seems to have no idea what you can and cannot include in a trademark application, or at what stage in the trademark process you should file an argument or petition against an existing trademark registration.

Chatbots tend to make irrelevant suggestions for new trademark applications like “argue that MARK A is in a different industry from MARK B and has a different appearance.” You cannot include a legal argument in your trademark application. You can sometimes file it in a way that implies an argument, but that’s not what ChatGPT tells my clients to tell me to do. 

I also frequently have existing clients ask me during the filing process to “confirm that the Safe package includes responses to major office actions,” sometimes several times in the same email thread. Why would somebody ever ask such a question? That’s the whole point of the Safe package and it’s the only difference between it and the Value package. If you need confirmation, it’s confirmed in several places that you’ve already seen, including the beginning of the very email thread you’re replying to. It’s why you paid more money! 

It’s very obvious that these clients are plugging my correspondence into a chatbot and then sending me the response without even reading or editing it first.

I also had a client ask multiple times for me to file the $250 “TEAS Plus” application type instead of the $350 “TEAS Standard” application type, even after I had already told him that the USPTO retired the entire TEAS system almost a year prior. The amount of worthless detritus that makes into into emails sent by these clients is indefensible. 

Can you imagine if every time a business colleague called you, they added the same extra two minutes of meaningless gibberish, scattered throughout the conversation? Why would that kind of behavior be okay simply because it’s in email form? It still wastes a similar amount of time.

So if you’re reading this and you’re thinking of becoming my client, or you’re simply thinking of dumping some slop on a coworker, hopefully you’ll think twice now.