Introduction
If I had to name just one reason why ADHD task management often fails, it is that “working memory (the ability to temporarily hold information for processing) drops information much faster than expected.” This article explores how to use AI to offload that information, delving into the level of prompt design.
Since the actual method of task decomposition is covered in another article, I will focus here on the technical aspect of “how to instruct AI so that it functions as a substitute for memory.”
What is working memory, and how does it break down with ADHD?
Working memory is the function that holds information related to what you are currently doing in your head for a short period. For example, tasks that require holding multiple pieces of information simultaneously, such as “remembering a deadline you were just told while working on a document,” fall under this category.
In ADHD, the time and amount of information that can be held tends to be short, regardless of conscious effort. As a result, it is easy to fall into a state where “what is not visible is the same as not existing,” and while working on the task in front of you, other plans and promises completely slip your mind. This is not laziness, but a characteristic of the mechanism for holding information itself.
Basic design for externalization
The basic strategy for this characteristic is to stop “remembering in your head” and switch to “putting it in a visible place,” which is known as externalization. Paper notes and calendars are also a form of externalization, but when using AI, you can go a step further and create a state where “the AI reorganizes the information you have put out on your behalf.”
There are three conditions for making externalization work.
-
Decide on one place to put information: If you scatter notes across multiple destinations, the externalized information itself becomes a burden on your working memory to remember “where you wrote it.”
-
Do not organize when putting it out: Write things down in the order they come to mind, and leave the organizing to the AI. If you try to organize while writing, the act of writing itself becomes heavy.
-
Design it so that you always look back at it at least once after putting it out: If you just put it out without looking back, the externalized information will not function as a substitute for working memory.
Common causes when externalization does not go well
If you have created an externalization mechanism but it is not working, the cause is often that one of the three basic design conditions has collapsed.
The place to put information has effectively become two or more locations
Cases where the person thinks they have decided on one place, but in reality, the places are scattered—such as “I basically put it in the chat, but I also write it in my smartphone notes when I think of it”—are common. In this case, start by looking back at where you wrote what for one week and recounting the places you are actually using.
You are organizing it first
If you try to organize in your head before putting it out, such as “I’ll make a bulleted list before I send it” or “I’ll arrange them in order of importance,” that organizing work itself puts a load on your working memory, and the meaning of externalization is diminished. Consciously allow yourself to send it in the order you thought of it, without organizing it.
The habit of looking back is not built into the design
If you keep putting out new information without ever looking back at the externalized information, the information will just accumulate without being utilized. You need to incorporate the act of looking back into the mechanism itself, such as by fixing the timing for looking back in your calendar or reminders, or by using the weekly inventory prompt introduced next.
Once you can identify which of these three is breaking down, you can often restore functionality by fixing just that one point, rather than having to overhaul the entire externalization process.
What changes when you delegate externalization to AI?
Paper notes and standard smartphone memo apps are also forms of externalization, but the difference when using AI is that you don’t have to be involved in ‘how to handle the information you’ve put out.’ Paper notes only hold the information you’ve written in its original form. When you look back at them later, you still have to use your own brain to sort out which tasks are for today and which are appointments for next month. In other words, even if you think you’ve externalized, the burden of sorting remains in your head.
When you output to AI, you can delegate the sorting task itself. Even if the information is completely unorganized at the moment you write it down, categories like ‘Things to do today,’ ‘Things I don’t need to think about today,’ and ‘Needs confirmation’ only emerge when you ask for it to be organized. Because the timing of outputting information and the timing of organizing it are separated, you don’t have to think about classification at all while you are writing.
Let’s look at how this difference plays out in actual practice with a common scenario. Suppose you have eight pieces of information in your head at the start of the day: five tasks you want to finish today, two tasks you only need to decide on by next week, and one task that cannot proceed without confirming with someone else. If you write them down as-is on a paper note, the eight tasks are just listed in the order you wrote them, and it’s not immediately obvious which are urgent. As a result, it’s easy to fall into a pattern of panicking in the afternoon when you remember a task that was supposed to be done today.
If you output the same eight items to AI, you can get a list sorted by urgency just by running a dedicated organization prompt immediately after writing them down. Furthermore, for tasks where information is missing (like the one in the example above that requires confirmation), the AI will suggest what needs to be confirmed at the same time it classifies the task. This ability to have missing points pointed out while being classified—something that doesn’t happen with paper notes—is the real difference in externalization using AI.
This separation works well because the AI acts as a receiver that doesn’t evaluate or judge anything during the ‘output’ phase. When you consult a person, they inevitably return reactions or opinions while you are still talking, which in itself becomes a factor that interrupts your thinking. If you fix the role in advance with a prompt, you can mechanically ensure that the AI keeps its promise not to interject with evaluations until you have finished outputting everything.
Another advantage of this mechanism is that you don’t need to think about the prompt from scratch every time. If you save the two types of ‘output’ and ‘organize’ templates once, you only need to call them up the next time you use them, which lowers the hurdle to starting the externalization process itself. As for where to save them, if you follow the same logic as the previously mentioned ‘decide on one place to output,’ keeping them in one place will prevent you from getting lost.
Three Basic Principles of Prompt Design
To make AI instructions (prompts) function as a substitute for working memory, keep the following three principles in mind.
Principle 1: Separate the ‘output’ prompt from the ‘organize’ prompt
If you ask in a single prompt to ‘write down everything you’ve thought of and prioritize it at the same time,’ the tasks of writing and judging occur simultaneously, which undermines the purpose of externalization (outputting without organizing). Use a prompt just for outputting first, and ask for organization at a different time.
Principle 2: Design the AI to ‘ask back’
Even if there is missing information in what you’ve written down, it can be difficult to notice it yourself. If you include a sentence in your prompt like ‘If information is missing, please ask me questions before organizing,’ the AI will ask you for the necessary information, making it easier to notice oversights.
Principle 3: Fix the output format
If the organization results come back in a different format every time, the burden of reviewing them increases. If you fix the output format to be the same every time, such as ‘Please output with three headings: Things to do today / Things I don’t need to think about today / Needs confirmation,’ the task of reviewing becomes faster.
In the following sections, I will provide the full text of prompts that actually incorporate the three principles above, how to use custom instructions and memory functions, and even a prompt for a weekly inventory check.
What you will learn in the following sections:
-
Full templates for ‘output’ and ‘organize’ dedicated prompts
-
How to distribute content to write for custom instructions versus memory functions
-
How to write specific sentences that make asking for clarification effective
-
Prompts for weekly reviews and organization rules for when information keeps increasing
