AI Study Planners for College: What AI Should and Shouldn't Do

When generative AI first became widely available, much of the conversation in education focused on one question: How do we stop students from using it to cheat?

That was a reasonable concern, but I have always thought it was too narrow. AI can obviously be used to do academic work a student is supposed to do, and we should take that seriously. At the same time, college students do an enormous amount of work around learning that is not actually the learning itself. They organize assignments, interpret schedules, estimate time, decide what to do first, break large projects into steps, keep track of goals, and continually adjust when their plans change.

Those are executive functioning demands, and I think they create a much more interesting question for AI: can we use the technology to help students manage the work without having it do the work for them?

That distinction has shaped much of how we have thought about Tanglio.

The Problem With Asking AI to Do the Learning

If a professor asks a student to write an essay because the act of developing an argument and communicating it in writing is part of the learning objective, having AI produce the essay creates an obvious problem. The student may end up with a finished product without developing the skill the assignment was designed to teach.

The same concern applies to other kinds of academic work. If the purpose of a problem set is to develop mathematical reasoning, getting the answers is not the same as developing the reasoning. If the purpose of reading is to wrestle with a difficult argument, an AI summary may give the student the conclusion while removing the process that was supposed to help them understand it.

This is not an argument against AI. It is an argument for asking what the learning objective is before deciding what we should automate.

UNESCO's guidance on generative AI in education emphasizes a human-centered approach, including human agency and meaningful human engagement in the use of AI. That seems like a sensible starting place. Technology should help students become more capable learners, not simply make it easier to bypass learning.

There Is a Lot of Work Around the Learning

Now imagine a different use of AI.

A student has a ten-page paper due in two weeks. The student understands the topic and is perfectly capable of doing the research and writing, but they are overwhelmed by the size of the assignment and have no idea where to begin. AI helps them identify a sequence: understand the prompt, select a topic, find sources, read and take notes, develop a thesis, outline, draft, revise, and check citations.

The student still does the research. The student still evaluates the sources. The student still develops the argument and writes the paper. What AI has done is help organize the process surrounding the intellectual work.

I see a meaningful difference between those two uses.

College students are constantly asked to perform executive functioning tasks that are necessary for academic success but are not themselves the academic objective. A chemistry professor wants a student to learn chemistry. The professor probably does not care whether the student manually copies every due date from a syllabus into a calendar. A history professor wants a student to analyze history. The educational value of the assignment does not depend on whether the student independently remembers that the first draft should probably be started before midnight on the due date.

We should be careful not to confuse unnecessary friction with rigor.

AI Can Be Useful Before the Student Starts

One of the most promising uses of AI is helping students get from an intention to an action.

"Work on paper" is not a very useful instruction. "Find three scholarly sources related to your topic" is much easier to begin. "Study for exam" is vague. "Review chapters four and five, identify the concepts you cannot explain without your notes, and spend thirty minutes practicing those first" gives the student a starting point.

For students with executive functioning difficulties, this matters because the barrier is often not understanding the importance of the work. The barrier is translating a large or ambiguous responsibility into something actionable.

AI can help with that translation. It can help a student break down a project, think through a realistic sequence, compare competing priorities, or create a first version of a plan that the student can then adjust.

The student remains responsible for deciding whether the plan makes sense.

AI Can Also Be Wrong

This last part matters because generative AI has an unfortunate ability to sound confident even when it is mistaken.

An AI planner may misunderstand an assignment, estimate time poorly, miss a nuance in a syllabus, or recommend a sequence that does not fit the student. It may also know nothing about the professor's expectations beyond the information it has been given.

Students therefore need to treat AI-generated plans as suggestions rather than instructions. Does this breakdown match the actual assignment? Is the deadline correct? Does this amount of work fit the time available? Is the student allowed to use AI for this purpose under the course or university policy?

AI literacy includes knowing when not to trust the AI. This is another reason I do not like the idea of technology quietly taking over executive functioning. Students need to remain part of the decision-making process because evaluating and adjusting the plan is itself an important skill.

The Goal Should Not Be to Eliminate Struggle

Education has an uncomfortable relationship with struggle. Some struggle is pointless. A student manually transferring thirty dates from five syllabi into a calendar may not learn anything particularly valuable from the process. Other struggle is exactly where the learning happens.

Writing is difficult partly because organizing ideas is difficult. Solving a complex problem is difficult because the student has to reason through uncertainty. Reading challenging material requires sustained attention and interpretation. If AI removes all of that struggle, it may also remove the thing we wanted the student to learn.

The question I would ask is not, "Can AI do this?" It almost certainly can do more of it every year. The better question is, "Should the student be doing this part?"

If the answer is yes because the process is central to the learning objective, AI should be used cautiously or not at all. If the answer is no because the task is administrative, organizational, or unnecessarily burdensome, AI may be exactly the right tool.

There will be plenty of gray area between those two categories, which is why universities, faculty, students, and technology companies need to keep having this conversation rather than pretending a single rule will solve it.

Why We Built Tanglio Around Executive Functioning

This distinction is central to Tanglio. We did not set out to build an AI system that writes students' papers, solves their homework, or replaces the academic work they came to college to do.

We built it around the work that surrounds learning.

A student can upload a syllabus so assignments and deadlines become easier to manage. They can look across their responsibilities and think about what deserves attention. A large assignment can be broken into smaller steps. Goals can be made visible. A day can become a plan rather than a collection of things the student vaguely hopes to remember.

Even the AI Notetaker should be understood through that lens. The goal is not to make attending class unnecessary. It is to reduce the burden of trying to capture everything while also listening, processing, and participating. Students still need to engage with the material and decide what matters.

I think this is where AI has enormous potential for students who struggle with executive functioning. It can provide structure at the moment the student needs it without requiring another person to be sitting next to them. But the measure of success should not be how much thinking the technology can eliminate. It should be whether the student is better able to direct their own thinking and work.

A Better Test for Educational AI

As AI becomes embedded in more educational tools, I think students, families, educators, and universities need a better test than whether something is technically impressive.

Ask what the technology is doing for the student. Is it helping them see information they already have? Is it helping them organize responsibilities, think through priorities, or break an overwhelming task into a manageable starting point? Is it prompting reflection and decision-making? Or is it simply producing the answer, paper, or product the student was supposed to learn how to create?

Those uses are not equivalent.

We are going to get some of this wrong. Every major technology changes the way we work, and education usually needs time to figure out which changes are useful and which ones undermine what we were trying to accomplish in the first place.

I am optimistic about AI in education precisely because I do not think its best use is replacing students. I think its best use may be helping students manage the increasingly complicated environment in which we expect them to learn.

If we can use AI to reduce unnecessary executive functioning barriers while preserving the thinking, effort, curiosity, and struggle that actually produce learning, that seems like a pretty good place to start.