How to Automate Meeting Notes for a Small Agency

Step 1: Record the Meeting

Everything downstream depends on a clean recording. Whether you're running calls on Zoom, Google Meet, or a phone line, you need a consistent recording setup that captures audio clearly and doesn't require manual steps mid-call.

The simplest approach is a meeting bot — a tool that joins your calls automatically, records the audio, and stores it for processing. Set it once and it handles every future meeting without extra steps. If you prefer privacy-first options, some tools run locally and record without joining as a bot participant.

Always notify participants at the start of the call that it's being recorded. This is both a legal requirement in many jurisdictions and a basic professional courtesy.

Step 2: Transcribe the Audio

A recording is only useful once it's searchable. Transcription converts your audio into text you can feed into AI tools, store in your project workspace, and reference later without replaying the whole call.

Most AI meeting tools handle transcription automatically as part of their recording workflow — you don't have to do a separate step. If you're working with existing recordings or using a tool that only records, you'll need a dedicated transcription step. Look for tools that support speaker identification so the transcript shows who said what.

Transcription accuracy varies by audio quality and technical vocabulary. After a few calls, you'll see where errors cluster — usually names, product terms, and acronyms. Some tools let you add a custom vocabulary to fix recurring mistakes.

Step 3: Summarize the Meeting

A full transcript is rarely what you need day-to-day. A good summary distills a 45-minute call into a structured two-page document: what was discussed, what was decided, and what still needs resolution.

Use an AI writing tool with a structured prompt that takes the transcript and outputs a consistent summary format. Define the sections you want — context, key discussion points, decisions made, open questions — and use the same template every time. Consistency makes summaries easy to scan and file.

Some meeting tools generate summaries automatically as part of the post-call workflow. These are good starting points but often benefit from a custom prompt that matches your agency's format.

Step 4: Extract Action Items with AI

The summary tells you what happened. Action items tell you what to do next. These need to be extracted separately — a combined document buries tasks inside narrative and they get missed.

Run a second prompt against the transcript or summary that pulls out every task, decision, and commitment made during the call. For each one, capture the owner, the deliverable, and a deadline if one was stated. Feed this list directly into your project management tool.

The best setups automate this step: the meeting ends, the transcript is processed, and tasks appear in your project board without anyone typing them manually. This is achievable with a simple Make or Zapier workflow connecting your transcription tool to ClickUp, Asana, or Notion.

You are an operations assistant for a small agency. I will paste the transcript of a client or internal meeting below.

Your task:
1. Extract every action item mentioned — whether assigned explicitly or implied by context.
2. For each action item, identify: the task, the owner (person responsible), and the deadline if mentioned.
3. Format the output as a numbered list. Use this structure for each item:
   - Task: [what needs to be done]
   - Owner: [who is responsible — use first name only]
   - Due: [deadline if stated, or "Not specified"]

Also note any decisions made during the meeting that do not require an action but should be documented.

Do not include discussion points that did not result in an action or decision. Be specific — avoid vague items like "follow up on project."

Transcript:
[paste transcript here]

Step 5: Draft the Follow-up Email

A follow-up email sent within 30 minutes of a client meeting signals professionalism and closes the loop before context fades. It should recap what was decided, confirm the next steps on both sides, and set a clear expectation for what happens next.

Use the action items from Step 4 as input. The prompt below generates a client-ready follow-up email that sounds professional without sounding robotic.

Always review before sending. The AI draft handles structure and completeness; you review for tone, accuracy, and anything sensitive that shouldn't go in writing. This review step should take under two minutes.

You are writing a follow-up email on behalf of a small agency account manager, to be sent to the client within 30 minutes of a meeting.

Use the action items and meeting notes below as your source. Write in a clear, professional, and warm tone — not overly formal, not casual.

Structure the email as follows:
1. One-sentence opener thanking the client for their time and naming the meeting topic.
2. A short paragraph summarising the two or three most important things discussed or decided.
3. A section titled "Next steps" listing what the agency will do and what (if anything) the client needs to do, with owners and deadlines where known.
4. A one-line closing that states when the client can expect the next update.

Do not use filler phrases like "As per our conversation" or "Please do not hesitate to contact me." Write naturally.

Meeting notes and action items:
[paste action items from Step 4 here]

Step 6: Generate the Internal Summary

The follow-up email is for the client. The internal summary is for your team. It captures context and nuance that doesn't belong in a client-facing document — project risks, client sentiment, internal dependencies, and things to watch.

Save this directly to the client's project folder. It becomes the reference point for anyone who wasn't on the call and the record you'll thank yourself for having in three months.

You are writing an internal meeting summary for a small agency's project file. This document is not shared with the client.

Using the transcript below, produce a structured internal summary with the following sections:

**Meeting overview**
Date, attendees, meeting type (client call, internal, kickoff, etc.), and a one-sentence purpose.

**Key decisions**
Bullet list of decisions made during the meeting that affect the project scope, timeline, or deliverables.

**Action items**
Bullet list of all tasks, with owner and deadline. Include both agency-side and client-side tasks.

**Risks and watch points**
Any concerns, blockers, or unclear areas that the team should monitor. Note if the client seemed uncertain or if scope is at risk of expanding.

**Client sentiment**
One to two sentences on how the client appeared during the call — confident, concerned, uncertain, enthusiastic. This helps the account manager prepare for the next interaction.

Be direct and factual. Write for an internal team reader who needs to get up to speed quickly.

Transcript:
[paste transcript here]

Common mistakes when automating meeting notes

1. Skipping the recording consent step — always notify participants. Failing to do so creates legal exposure and erodes trust.
2. Using a generic summary prompt — a prompt that asks for “key points” produces generic output. Define the exact sections you want every time.
3. Combining summary and action items in one pass — the AI will prioritize narrative over tasks. Run them as separate prompts for cleaner results.
4. Not reviewing before sending — automated follow-ups sent without review will eventually contain errors or tone mismatches. Build in a two-minute review every time.
5. Storing transcripts without a clear retention policy — transcripts contain sensitive client information. Decide upfront how long you keep them and where.

Simple stack for this workflow

Related resources

What to do next