In brief. Use a short record of planned and completed work, ask AI to help you spot questions, and choose one small adjustment to test. Keep observations separate from explanations and retain responsibility for your priorities.
A weekly work review should leave you with a clearer next week, not a flattering account of the last one. An AI coach can help organise your notes and ask follow-up questions, but it cannot know why a task slipped unless the evidence is there. The most useful starting point is a small record of what you intended, what happened and what remains uncertain.
This guide describes an original planning routine for your own work. It makes no claim that a chatbot can objectively assess your performance or predict your productivity. NIST’s generative AI profile discusses confabulation and human–AI configuration risks. Those concerns are a reason to check the story a tool builds around your week rather than accepting it as an expert verdict.
Prepare a short evidence packet
Write down three things: the tasks you planned to finish, the outcomes you can confirm and the commitments that changed during the week. Use plain descriptions such as “sent the draft on Thursday” or “received the missing figures on Friday.” If you estimated time, distinguish your original estimate from the time you actually recorded. Do not fill gaps with numbers that merely feel plausible.
Keep the packet limited to information you are allowed to share with the tool. You can describe a delayed approval without naming a customer or pasting a private message. Include only the detail needed for the review. KINSMC’s guide to checking how an AI tool can use your data covers the separate question of data settings and service terms.
Review your own explanation first
Before opening the chat, write two sentences about what helped and what got in the way. This gives you something to compare with the coach’s interpretation. You may discover that you blamed interruptions while the record shows an unexpectedly large revision. Alternatively, your explanation may contain context missing from the task list. Neither your initial account nor the chatbot’s account needs to win automatically.
For example, imagine planning a short internal newsletter and finishing it two days late. The record shows that new figures arrived after the planned drafting slot. That supports asking how to handle late inputs next time. It does not, by itself, establish that you procrastinate, that another person was careless or that the whole workflow needs replacing.
Ask for questions before recommendations
Use a prompt such as: “Review these notes about my own week. Separate observations from possible explanations. Ask up to three questions where context is missing. Do not infer motives or invent time estimates.” Answer only the questions that help clarify a real planning issue. You can reject an assumption or say that the information was not recorded.
Then ask for a small number of possible adjustments, each tied to an observation. A suggestion to reserve a drafting slot should explain what problem that slot is meant to address. A suggestion to reduce meeting time needs evidence that meetings displaced the work in question. If a recommendation cannot point to an observation, keep it as an idea to investigate rather than an action you have agreed to take.
Choose one adjustment you can observe
Select a change small enough to try during the next week. For the newsletter, you might request the figures by Tuesday and draft the non-numerical sections while waiting. Write down what you expect to notice: whether the draft is ready before the figures arrive, or whether missing inputs still block most of the work. Avoid promising a percentage improvement without a basis for measuring it.
A useful experiment includes a change, a review date and a condition for reconsidering it. You might decide to try the new sequence once, then ask whether it reduced waiting or merely moved the delay elsewhere. Keep a brief decision log if the adjustment affects an ongoing team process. Confirm any change to another person’s commitments with them directly.
Finish with a plan you actually own
Write the next week’s priorities yourself after reading the suggestions. Give each priority a concrete next action and note any dependency. “Prepare the outline before Tuesday’s update” is actionable; “be more strategic” is not. Check the plan against your existing commitments and available time before accepting new work. The coach’s ability to generate a longer list does not create more capacity.
Keep the review output to one short page: completed work, unfinished work, important context, the adjustment you chose and the next review question. Save the evidence behind any consequential claim. A compact record is easier to revisit than a full transcript containing many suggestions you never accepted.
Check the adjustment at the next review
Begin the following review by looking at the previous experiment. Did the change happen? What result can you observe? What else changed that week? If the newsletter was early during a quieter week, you cannot attribute the whole difference to the new sequence. Record that uncertainty instead of asking the AI to turn one week into a confident causal explanation.
Continue, revise or stop the adjustment based on what you learned. The purpose of the coach is to help you ask better questions about your work and translate your decisions into a usable plan. The evidence, the priorities and the decision to act remain yours.
