How to extract meeting summaries: decisions, tasks, and open questions
The Meeting Summary tool helps quickly extract decisions, open questions, and follow-up actions from meeting text. It works with text only: you paste a transcript or notes, optionally specify a goal, and receive a single result that can be copied or downloaded as a plain text TXT file. In this guide we will walk through how to prepare your source material, how to formulate your goal to get the right focus, and how to check the result for completeness and accuracy. We do not guarantee flawlessness: the model may make errors, and service availability depends on queueing. All examples in this article are fictional and provided only for illustration.
Open tool: Meeting summary ↗What the tool accepts and in what form
At input there are two text fields. "Source material" is the meeting text: a transcript, notes, or a brief summary. Maximum volume is 8,000 characters. "Goal or desired result" is up to 1,000 characters; here you describe what to focus on. The tool does not process audio or video, so any recording must first be converted to text manually or in another service, then pasted into the field. There are no settings, styles, change levels, or additional panels in the tool.
The result is output once in a separate panel. It can be copied or downloaded as a plain text TXT file. No other export formats are available. The tool does not save results automatically, does not email them, and does not publish them. If a meeting is long, split the text into parts and process each separately, then combine the summaries manually.
It is useful to pre-process the text before running the tool: remove duplicates, expand abbreviations, and explicitly name participants. The clearer the source material, the fewer opportunities for errors. Remember that the model may miss important points or add extraneous ones, so always check the output against the original text.
Source material: "Anna will prepare a draft by Friday. Sam will check it on Monday. The launch date remains unresolved." Goal: "Extract tasks and responsible persons."
How to formulate a goal to get the right focus
The goal directs which text fragments the tool highlights. If you write "extract only decisions," the result will lean toward final agreements. If you write "extract tasks with responsible persons and deadlines," the focus shifts to assignments. Formulate your goal specifically: "list decisions," "show tasks and responsible persons," "name unresolved questions." Avoid vague requests like "do it well" — they do not help the tool.
You can specify the output language if it differs from the source text language. However, translation quality may vary, so verify the formulations. Do not ask the tool to set priorities, give recommendations, or evaluate the discussion quality: it only extracts information from the text.
If the first result is not to your liking, change the goal and run the tool again. This is the main way to get a different summary variant. Do not try to achieve the desired output through additional settings — they are not available.
Goal: "Only decisions, no tasks and open questions." Result: "Decision: project launch postponed until budget approval." (Fictional example.)
Decisions: how to recognize and verify them
Decisions are final agreements. In text they often appear as phrases "we decided," "approved," "confirmed." If your goal says "only decisions," the tool will try to highlight exactly such statements. This is convenient when you need to quickly fix key points for a report or protocol.
However, the tool may mistake a discussion or a proposal for a decision. Check every item: is it really an agreement, or just an option? Conditional formulations like "possibly," "if," "may be" are usually not decisions. Compare the output with the source text and remove anything superfluous.
If the text contained no explicit decisions, do not expect the tool to invent them. In such a case it is better to honestly note that no decisions were made and move the discussion to open questions.
Source text: "Possibly we will postpone the launch." The tool may erroneously output this as a decision. Verification: this is not a decision, but a proposal.
Tasks and responsible persons: what is extracted, what is not
Tasks are assignments with an action and, preferably, an executor. If your goal says "extract tasks, responsible persons, and deadlines," the tool will find sentences with action verbs and names. For example, "Anna will prepare a draft by Friday" becomes a task: responsible — Anna, action — prepare draft, deadline — Friday. If no deadline is mentioned, the tool may not add one.
The tool does not verify real deadlines, does not set priorities, and does not assign responsible persons. If the text says "we should check the report," a task may not be recognized due to the lack of an explicit executor and an imperative verb. In such a case add the formulation manually or clarify the source text.
Check that tasks mentioned in passing are not lost. Cross-reference the list with the source material and supplement if necessary. Remember: the tool does not replace a coordinator; it only helps not forget assignments.
Goal: "Tasks with responsible persons." Result: "Anna — prepare draft by Friday. Sam — check draft on Monday." (Fictional example.)
Open questions: how not to lose the unresolved
Open questions are topics that remained without a final answer. To get them, specify in your goal: "list unresolved questions and topics requiring further discussion." The tool will highlight sentences with a questioning intonation or phrases like "unresolved," "needs clarification," "let us discuss later." This helps plan the next meeting and not miss important items.
The tool may not recognize a question if it is formulated as a statement. For example, "We do not know when to launch" is an open question, but without a question mark. Check that all uncertainties are included and add any missed items manually.
Do not expect the tool to propose a solution for an open question. It only records that a question remained open. The solution remains with the meeting participants.
Goal: "Open questions." Result: "Launch date remains unresolved." (Fictional example.)
Comparing variants and manual fine-tuning
After receiving several variants with different goals, compare them for completeness and accuracy. Start from the source text: are all key points reflected? Are there any extraneous items not present in the meeting? Is the meaning not distorted? For example, if the text says "possibly postpone" and the decision reads "we postponed," that is an error.
Choose the variant that best suits your purpose: for a management report, decisions are more important; for team coordination, tasks; for planning the next meeting, open questions. You can combine the best parts from different variants into one document manually.
After choosing, verify names and deadlines, bring formulations to a consistent style, add the meeting date and participants if they were not in the text. The tool does not do this automatically. The finished document can be copied or downloaded as TXT; other export formats are not available.
Comparison: the decisions variant has 2 items, the tasks variant has 3, the open questions variant has 1. Choose the one without gaps.
Limitations and typical errors
The tool does not recognize speech and does not process audio or video. If you have a recording, first obtain the text. Also, it does not generate images and does not create audio: vector images and speech are separate functions unrelated to meeting summaries. Do not confuse them with the main task.
The model may make errors: miss important points, add non-existent ones, incorrectly identify responsible persons. Always check the result against the source text. Do not blindly trust and do not use the output as the sole source of truth.
Do not expect the tool to set priorities, give recommendations, or evaluate the discussion quality. If the text lacks clear formulations, the result will be blurred. In such a case rephrase the source text or split it into parts.
Error: the tool did not recognize the task "check the report" because there was no explicit executor. Solution: add "Sam will check the report" to the text.
The Meeting Summary tool helps extract decisions, tasks, and open questions from text. To get different focuses, change the goal formulation and run the tool again. Compare variants, check them against the source text, and manually fine-tune. Remember the limitations: text only, up to 8,000 characters, no audio or video, possible errors, availability depends on queue. Examples in this article are fictional. Use the tool as an assistant, but responsibility for accuracy of summaries rests with you.
Open tool: Meeting summary ↗Interface translations are machine-assisted and may need refinement.