Start with Your Raw Material
The first step is the simplest: get your notes into a digital format. Don't worry about organizing them. Whether it's a rambling voice memo transcript, a scattered collection of bullet points from a brainstorm, or a copy-pasted document full of research
highlights, Claude can handle the mess. You can get a transcript from services like Zoom, Teams, or Otter.ai, or simply type up your handwritten scrawls. The key is not to edit or censor yourself. The more raw context you provide the AI, the more effectively it can identify the underlying structure, commitments, and key ideas that are currently buried in the chaos.
Use a Role-Playing Prompt
Treating Claude like a search engine yields generic results. Instead, give it a role. Think of it as a highly competent executive assistant or chief of staff. This framing helps the AI adopt the right mindset for the task. Your prompt is the most critical part of this process. Start by setting the scene. A great prompt structure includes the role, the context of your notes, and a very specific task. The goal is to move beyond simple summarization, which is often easy to ignore. Here’s a powerful template to start with: 'You are an expert executive assistant responsible for creating clarity. Below are my messy, raw notes from a recent project meeting. Your task is to analyze them and structure the information into a useful format. [PASTE YOUR NOTES HERE]'
Extract Key Decisions and Open Questions
An action plan is more than just a to_do list. Before identifying tasks, you need to know what was actually decided. After providing the role and context, ask Claude to pull out two specific things: key decisions and unresolved questions. A decision is a commitment that has been made, which provides context for future actions. An open question is a topic that was raised but not settled, which often represents a hidden risk or a future agenda item. Ask for these in distinct sections. For example: 'First, create a bulleted list under the heading 'Key Decisions Made'. Then, create a separate list under the heading 'Open Questions for Follow_Up'.'
Generate a Specific Action Table
This is where the magic happens. A vague list of 'next steps' is not an action plan. An effective action item has three components: a specific verb, an owner, and a due date. Anything less is just a topic of conversation. To get a truly useful output, instruct Claude to create a table. Most importantly, tell it not to guess. If the owner or due date wasn't mentioned in the notes, it should explicitly state 'UNCLEAR' or 'Not Specified'. This prevents the AI from 'hallucinating' commitments that were never made. Your prompt should be explicit: 'Finally, create a table with three columns: 'Action Item', 'Owner', and 'Due Date'. Only include firm commitments made in the notes. If an owner or date is not mentioned, write 'Not Specified' in the corresponding cell.'
Refine and Iterate
The first output from Claude is a powerful first draft, not necessarily the final product. Now you can refine it. Want the action plan in a different format? Ask for it. For example: 'Now, take that action table and draft a professional follow_up email to the team, summarizing the decisions and clearly listing each person's responsibilities.' You can also ask for different views based on the audience. For a client meeting, you might ask Claude to rephrase the summary using plain language, focusing only on mutual commitments. This iterative process of conversation allows you to quickly shape the raw information into precisely the format you need, saving hours of manual editing and re_typing.
