Problem Solving and Planning

Thinking Aid, Not Decision-Maker

AI is useful for thinking through a problem and unreliable for deciding what to do about it. That distinction carries most of the weight on this page. Used as a thinking aid it will surface options you had not considered and expose assumptions you did not know you were making. Used as a decision-maker it produces confident recommendations built on nothing it actually knows about your situation.

What It Is Good At

Using AI as a starting point for a brainstorming session can actually hinder your ability to find credible solutions and even hurt the needed skill of critical thinking since you are offloading that thinking to the AI. Use it or lose it. Also, if everyone uses AI to come up with ideas, the ideas will start to be very similar in nature and there won't be any differentiation in your work.

A better technique that will result in more applicable ideas and continue to reinforce your critical thinking skills is to ask the AI to conduct an interview with you. This allows the AI to pull out your own ideas and thoughts about the topic to help you frame them. Once you have a clear picture of the problem statement and potential solutions you think could be good, you can explore gaps and ideas with the AI that you hadn't considered previously and evaluate the ideas against each other. 

The most common failure in planning is not choosing badly among options; it is never generating enough options to choose from. Groups converge fast on the first plausible idea.

AI is good at the widening step. "Give me eight approaches to this, including a few that are impractical." "What would this look like with no budget? With triple the budget?" "How would a completely different kind of organization handle this?"

Most of what comes back will be unusable. That is fine — the value is in the two you had not thought of, and in the way an obviously wrong option clarifies why the right one is right.

  • "What am I assuming here that I have not stated?"
  • "Who is affected by this who is not in the room?"
  • "What would have to be true for this plan to fail?"
  • "What is the strongest objection someone would raise?"

The pre-mortem is particularly worth doing: describe the plan, then ask the tool to imagine it is six months later and the plan has failed badly — what happened? This surfaces risks more effectively than asking "what are the risks," because it forces a specific story rather than a generic list.

Describe a project and ask for it decomposed into phases, tasks, and dependencies. You will get something roughly right that is much faster to correct than to build.

Useful follow-ups: "What is on the critical path?" "What can run in parallel?" "What does this plan assume someone else will do?" That last one catches the dependency nobody documented.

Expect timelines to be optimistic. These tools do not know your institution's approval cycles, your committee calendar, or how long anything actually takes here.

When you are weighing options, ask for a structured comparison: the criteria that matter, how each option performs, what each one costs you.

The value is less in the answers than in being forced to name the criteria. Teams frequently discover at this point that they disagree about what they are optimizing for — which is worth finding out before the decision rather than after.

"I need to explain this change to a group that will not like it. What will they push back on?" "What questions should I expect?" "How do I explain this to someone without the technical background?"

Useful for anticipating reactions and pressure-testing your framing. Not useful for writing what you will say — that has to be yours, or it will sound like it is not.

Where to Stop

The line is not about difficulty. It is about whether the decision requires knowledge of a specific situation and accountability for the outcome.

Do not use AI to decide:

  • Anything about a specific person — hiring, evaluation, discipline, advising, accommodation
  • Anything requiring current institutional knowledge — policy, budget, who has authority, what was decided last spring
  • Anything where being confidently wrong causes real harm
  • Anything you could not explain and defend on your own

Two quieter risks

Anchoring. Once you have seen the AI's recommendation, it is genuinely hard to think past it. If the decision matters, form your own view first, then use the tool to challenge it.

False authority in group settings. "The AI analysis says" carries a weight it has not earned, and it can shut down disagreement from people whose judgment is better informed. If AI contributed to something you bring to a meeting, say so plainly, and be clear about what it did and did not know.

Before anyone else depends on it

If you are planning something that will use AI as part of an ongoing process — an automated workflow, a tool embedded in a service, anything touching student or personnel data — that is a different kind of project with review requirements attached. Data handling, accessibility, procurement, and records retention obligations apply regardless of whether the tool is free.

Choosing the Right Tool covers what to check and who to talk to.