PROGRAMMING NOTE: If you don’t have time to read this whole article now (maybe have a bot read it for you), we are doing a management-only webinar on April 11, 2023, at 3 PM ET. Invites will go out soon, but if you want to be pre-registered, click this link.
You can’t open the news without seeing articles about OpenAI’s ChatGPT or its rival LaMDA from Google. We’ve reached an inflection point for Artificial Intelligence (AI) driven applications. More and more people are experimenting with these tools. You can expect them to improve exponentially.
Oh, and in case you’re wondering, that paragraph (and the rest of the following paragraphs) was written by PI (Phil Intelligence), which is lagging further and further behind AI :)
I’m not going to do a primer on AI. If you want to understand how these tools work, check out thiscomprehensive article or a simpler one here. Instead, let me share some of the things we’ve been playing around with and how I think AI-powered tools are likely to be employed in the Labor Relations world. I’ll also share some of the concerns I believe Labor professionals should be thinking about as these technologies develop and grow.
Is AI coming for your Labor Relations job?
First, the good news, I don’t think AI is coming for your Labor Relations job anytime soon. Tools like ChatGPT will radically transform how we work and how efficiently we can accomplish low-expertise tasks. But at its very center, Labor Relations is about relationships and complex people strategies and decisions. Based on how it handles the lower-level tasks today (which will improve quickly but is remedial), AI isn’t going to be running the core value Labor professionals provide in the world.
While your job isn’t going away, it is going to change. Here are some key areas we are exploring:
The research will get much more efficient. With Bing, you can ask specific questions (and iterate them) with links to the primary sources. This technology is a significant improvement over your typical Google search, which often requires pouring through dozens of pages (or dozens of inquiries) to get what you’re looking for, especially something off the beaten path. And based on our early tests, answers to many questions that would otherwise require digging through several pages of text just get answered.
SFDs are produced in seconds. Around here, we use the term SFD to refer to a Shi##y First Draft. Get me some words on a page that are a solid starting point for editing. ChatGPT is built for this task, and it’s only going to improve. Ask it to write a paragraph or an email draft about something specific; it does a credible job in seconds. We’ve given it some complicated tasks. While the output is objectively weak, it is a great head start for someone with a lot of experience, and it happens almost as quickly as you can ask the question.
Speaking of drafting, one area where these AI tools will have a major impact is legal drafting. I have some critical cautions below, but I’ve experimented with using the tools to draft contract language and policies, and it does a surprisingly good job for an SFD. Remember the S part of SFD—these aren’t clauses you can immediately put to work, and it requires an expert to tell the difference between something that looks good versus something that is good. But it’s a great way to get draft language for editing, and they do a decent job with pretty complex labor and employee relations topics.
Force multiply: One big problem I face, and I know many other Labor Pros feel the same way, is that it’s impossible to keep up with everything happening in our world. I have a team that helps me do this, and it’s still impossible. These new AI tools can help you force multiply in exciting ways. I can have a bot sit in on a meeting or watch a video for me and draft a summary of the key points covered. I can have it summarize web pages and articles. I have used RSS news feed aggregators for these tasks, but it’s easy to fall behind because you must read full articles (often repetitive) to find the nuggets. This is a way to shortcut a lot of that and point you to the most important things to spend time on reading.
Other tools: I’ve only dealt with text-based tools here, but there are many different tools around images, voice, video, and much more being developed every day. We are just scratching the surface.
Two Cautions
It’s all about the prompt: AI tools are still new, and we have noticed that the key to decent output is a decent “prompt” or the question you ask in the first place. Plus, these tools are designed with boundaries that could prevent them from answering your question, depending on your prompt. The critical skill for the future is writing the prompts. I predict that while AI isn’t necessarily going to destroy a bunch of lower-level legal and support roles, it is going to transform those roles into being great prompt writers. And, of course, AI tools are already designed for prompt writing. This will soon become an important part of training new Labor Pros.
Dunning-Kruger problem
If you’re unfamiliar with the Dunning-Kruger effect, it describes the fact that people who know nothing about a subject are likely to feel the most confident about their knowledge of the subject. As someone grows in knowledge and experience, they will learn how complex the subject is and get less confident in their knowledge. Eventually, they will become knowledgeable enough to gain confidence, but they will still be less confident than someone who knows nothing.
Which leads to this perplexing truth: The first rule of the Dunning-Kruger Club is you don’t know you’re in the Dunning-Kruger Club. That relates to AI tools. Take a look at this chart: