Three Courtesies We Are Giving AI (But Not Our People)

In my summertime, I am taking advantage of the time I took off to take some classes. One interesting subject I am all in on these days is, obviously, AI. One day, as I was listening to the instructor of one of my classes explaining the basic ground rules of prompting to obtain better results from our different AI tools, I had an AHA! moment. We are giving AI more courtesies than we give our people!

One of the things that our instructor mentioned is that all AI chats produce better results if they are given not only a task, but a task surrounded by relevant context and examples of what a good work product would be. In other words, we must tell the chats what we want, explain a little bit about what we want, and then provide an example of what we consider a good result.

Thanks for reading! Subscribe for free to receive new posts and support my work.

OMG! If we did the same for our humans, it would also lead to amazing results. I realized, while training on how to use my AI, that I was being a little bit of a prick by assuming my team would know what I mean when I ask for something, that they would know how I want it done, and also that they would read my mind on the standard of what is good and what is not good, in my eyes…

Wow, as I finished my class, I realized that there are three basic courtesies we are giving AI, but we do not give to our humans:

  1. Provide good context.

Instructions for tasks without context are useless. Even the smartest person cannot produce the result you expect if you do not give enough relevant context. Even for AI, which is hyper literal, it can be difficult to determine what you mean when you ask for something that may be interpreted differently across different fields of knowledge. You must tell it, you must give it a clear field of knowledge. Is it sports, business, education, marketing, research, or finances? An omniscient computer needs to be situated within a field of knowledge to be accurate. Why, then, do we expect fallible humans to understand the field of knowledge in which our thoughts are situated?

The art of crafting useful context is providing enough context while distilling the useful context from the garbage that is not relevant. How many times do we start talking to our employees about our weekend and then shift to instructions, then go back to the gossip, then run back to the instructions?

We expect employees to be able to understand what was garbage and what was useful from our “instructions”: plan the networking lunch with the networking group, dislike the gossipy lady with the tacky dress that you saw at church, who made a bad comment about your mom, but still invite her because she gives you good leads? … or not? Do you follow? Exactly!

  1. Compartmentalize each task’s information and forget instructions given in connection to other tasks.

New AI chats can be created every time we ask for something. If we want to review an email, the good practice on LLMs such as ChatGPT or Claude is to create a a new chat or task for each purpose so it will keep the context you give only for that task. There is a limited background data you can ask the LLM to keep on you so it is able to have a general idea of your personality, style, background etc. Nonetheless, as a general rule, each chat should contain the information that is specifically relevant to the task at hand. Too much background information can become noise and make it more difficult for the AI to identify and follow the instructions that matter most for that particular task.

In general, every chat is limited to the task it covers. If you open a new chat and reference the email you were reviewing in the other chat, the AI will not have a memory of it. This is done to avoid clutter and confusion. As instructions and information are provided, it becomes like a canvas full of paint: thick, inflexible, and burdensome. Therefore, the best practice is to provide a clean canvas to work on for every new task. That is amazing.

How many times do we tell our employees that we do not like a certain thing, and then when they avoid that same thing the next time, we get frustrated because we only meant it for that one time? We should be clear about what we are applying in general and what are one-timers. We should create our “background DNA” and then tell our team that except that DNA, for every task they should ask for background and assume it is a clean canvas. We should give our teams the courtesy of having a clean start and spend enough time creating good “prompts” for our peeps.

Most importantly, we should tell our team members that they are free to ask for explanations about exceptions to our rules. As business owners we need to be flexible, however, we should be able to accept questions when we change routes. Our teams are there in the trenches with us, so they understand if we need to change course, however, it is not fair to simply give them instructions in opposition to our normal actions and expect them to understand what to do now and what to do in the future under similar instructions. They cannot read our minds.

Finally, I we need to understand that it is not fair to be personally inconsistent, but expect good corporate habits. Either we invest the time and discipline required to build consistent habits and explain exceptions, or we are ok with caos and complete lack of uniformity. One or the other.

  1. Define the role for each task.

We tell AI what role it should play to improve its response. Stronger prompts often include both who we are and who the AI is. For example, if I am acting as the law partner at my business law firm, I might ask the AI to act as my marketing strategist for networking and relationship building. If I am researching a legal trend, I might ask it to act as my law clerk. That immediately sets the context and scope of the work and improves the quality of the response.

How often do we do the same for people? How often do you walk into the office and tell your assistant, “This morning we are focusing on marketing, so I need you to think like my marketing assistant”? As business owners, we wear many hats, and our teams do too, but, do we tell the team when to change one hat for the other, or, do we expect them to guess?

The closer someone works with us, the more roles they may need to switch between. I see that clearly with my assistant. On Fridays, she supports marketing. On Mondays, she helps coordinate clients. On Tuesdays, during our L10 team meetings, she helps track action items for the firm; hence, she is a saint!

She succeeds because we have set times for certain kinds of work. Still, when I change plans suddenly, I can confuse her about what role she is expected to play. That is something I am working to improve. If I tell my ChatGPT what it is supposed to be, I should have the same courtesy with my right-hand person.

As I study AI, I keep noticing the same lesson: we often leave people guessing about what we want from them and from the work. Are we focused on doing and we are not so great at communicating. Now, with these technologies that require our distilled input we are at least becoming aware of the need of clear instructions. The better we get at prompting AI, the better we should get at leading people.

On the other hand, as I study more AI, I am increasingly convinced that it is equally urgent to study human.

We will learn how to work with these new tools, but the most fundamental human truths will become the real differentiators. Teams may become smaller as AI takes over repetitive and predictive tasks. Yet people will still be responsible for navigating relationships, exercising intuition, managing uncertainty, and responding to variables that cannot always be anticipated.

These are fundamentally human responsibilities: communicating clearly, building trust, exercising judgment, and creating healthy working environments.

A strong place to begin is with a simple core value: never treat AI better than we treat our people. If we hold ourselves to that principle, we will be off to a good start.

In love and in the practice of law,

Your Bz Esq.

Contact us

Reserve your Appointment!

Complete this form so our team can contact you and reserve your appointment. You are welcome to share a few details about your legal needs so we can better understand your inquiry and connect you with the right attorney.

"*" indicates required fields

This field is for validation purposes and should be left unchanged.