Automate HR meeting summaries by connecting your cloud recording folder to Make.com, running audio through a transcription API, passing the transcript to an LLM for structured summarization, and routing the output directly into your HRIS—zero manual typing required, completed in under 10 minutes per meeting.

Every HR team runs on meetings. Interviews, performance reviews, disciplinary conversations, onboarding check-ins, strategy sessions—each one produces decisions and action items that need to be documented, distributed, and acted on. The problem is not discipline. The problem is that manual note synthesis is a slow, error-prone, and deeply low-value use of time that no one has bothered to automate yet.

This guide builds the workflow that changes that. Deterministic automation handles the mechanical spine first—file detection, API calls, data routing—and AI fires only at the discrete judgment point where rules cannot decide: summarizing language. Sequence is everything. If you want to understand the broader pattern before diving in, see why automation-first beats AI-first every time, and review the 7 questions to ask before you automate anything.

By the end of this guide you will have a production-ready Make.com scenario that detects a new meeting recording, transcribes it, generates a structured summary with decisions and action items, and pushes the output into your HRIS—in under 10 minutes, without anyone typing a single word. For teams newer to the platform, the plain-English guide to Make scenarios is a useful primer before Step 3.

What You Need Before You Start

This workflow requires four things in place before you open Make.com. Missing any one of them stalls implementation mid-build.

• Cloud storage with a dedicated recordings folder. Google Drive, Dropbox, OneDrive, or any S3-compatible storage. Your video conferencing platform must auto-save recordings to a specific folder within this storage.
• A transcription API account. OpenAI Whisper (via the OpenAI API) is the standard choice for teams already using GPT-class models. Google Cloud Speech-to-Text is the alternative for data-residency requirements. Have your API key ready before you build.
• An OpenAI API account (or equivalent LLM API). Whisper handles transcription; a separate GPT call handles summarization. They are different calls with different purposes—do not conflate them.
• Write access to your destination system. Whether that is an HRIS, an ATS, Notion, or a Google Sheet serving as a temporary knowledge base, confirm API credentials or Make.com native connector write permissions before building.
• Time to build. Initial build: 2–3 hours for someone new to Make.com. Refinement and testing: 1–2 additional hours. Error handling setup: 30–60 minutes. Total: one focused afternoon.
• A compliance check. Before processing real meeting recordings, confirm that your organization’s data retention policy, applicable employment law, and your transcription API’s data processing agreement are aligned. Disciplinary and accommodation-related conversations carry elevated legal sensitivity.

Step 1: Audit and Baseline Your Current Meeting Note Workflow

Before building, measure what you are replacing. You cannot prove ROI or catch edge cases without a baseline. Spend 30 minutes answering these questions:

• How many meetings per week produce notes or summaries that get filed anywhere?
• Who writes those notes—the facilitator, a dedicated note-taker, or the most junior person in the room?
• How long does summarization take per meeting, on average?
• Where do finished summaries go—HRIS, shared drive, email thread, or nowhere consistently?
• How often do action items from meetings go untracked?

This audit serves two purposes. First, it surfaces the meeting types and routing rules you need to encode in the workflow. Second, it creates the before-state data you need to calculate ROI after go-live. Research from Asana’s Anatomy of Work report finds that knowledge workers spend a disproportionate share of their workday on work about work—documentation, status updates, coordination overhead—rather than skilled work. Meeting summarization sits squarely in that category.

Log your baseline numbers. You will need them in Step 7. For a structured approach to this kind of discovery, see how to run an OpsMap™ audit before automating anything.

Step 2: Configure Cloud Storage and Recording Auto-Save

The workflow trigger is a new file appearing in a watched folder. Everything downstream depends on recordings landing there reliably and automatically—not on someone manually uploading a file.

Configure your video conferencing platform to auto-save cloud recordings:

• Zoom: Settings → Recording → Cloud Recording → set save location to your designated folder via Zoom’s cloud storage integration, or use Zoom’s built-in cloud recording and retrieve via API.
• Microsoft Teams: Meeting recordings save automatically to OneDrive or SharePoint depending on meeting type. Set a consistent destination folder through your Teams admin policy.
• Google Meet: Recordings save to the meeting organizer’s Google Drive. Use a shared Drive folder with a consistent path for HR meetings.

Create a folder structure that separates meeting types from the start: /HR-Recordings/Interviews/, /HR-Recordings/Performance-Reviews/, /HR-Recordings/Strategy/. Make.com can watch a parent folder and detect subfolders, which lets you route different meeting types to different summarization prompts and different HRIS destinations in a single scenario.

Test the auto-save by running a 2-minute test recording and confirming the file appears in the correct folder within 5 minutes of the meeting ending. Do not proceed to Step 3 until this is confirmed.

Step 3: Build the Make.com Watch Folder Trigger

Log into Make.com and create a new scenario. The first module is your trigger. If you want AI assistance building this, the walkthrough on building a Make scenario with Claude covers the same pattern.

1. Click the + to add the first module and search for your cloud storage provider (Google Drive, Dropbox, OneDrive, or Box).
2. Select the Watch Files (or Watch New Files in Folder) trigger module.
3. Authenticate with your cloud storage account using a dedicated service account—not a personal account. Service accounts prevent the workflow from breaking when individuals leave the organization.
4. Set the watched folder to your HR recordings parent folder path.
5. Set the trigger to activate on new files only, not on file modifications.
6. Add a Filter after the trigger to restrict processing to audio and video file extensions: .mp4, .m4a, .webm, .mp3. This prevents the scenario from firing on slide decks or PDFs accidentally dropped into the folder.
7. Set the schedule to run every 15 minutes. For high-volume HR teams, every 5 minutes is appropriate if your Make.com plan supports the operation count.

Save the scenario and click Run Once with a test file in the folder. Confirm the trigger fires and the file metadata (name, path, size) is visible in the output bundle before proceeding.

Step 4: Transcribe the Audio with OpenAI Whisper

The transcription step converts the raw audio file into text. This is a separate API call from the summarization step—keep them as distinct modules.

1. Add an HTTP module after the trigger. Make.com’s native OpenAI module handles Whisper as of 2025; use it if available on your plan, otherwise use the HTTP module with a direct API call to https://api.openai.com/v1/audio/transcriptions.
2. Set the method to POST and add your OpenAI API key as a Bearer token in the Authorization header.
3. In the body, set model to whisper-1 and map the file from the trigger module’s output as the file parameter.
4. Set response_format to text for a plain transcript, or verbose_json if you need timestamps for compliance documentation.
5. Add a Router module after the transcription step. Route by subfolder path so that interview recordings, performance review recordings, and strategy session recordings each go to a different summarization branch. Different meeting types require different prompt structures—this is where the folder structure from Step 2 pays off.

For meetings longer than 25 minutes, Whisper’s file size limit (25MB for the API) becomes relevant. Use the ffmpeg approach via a cloud function or a Make.com custom app to split long files before transcription, or use a streaming transcription service for recordings over 60 minutes. See how to feed API docs into Claude to build Make HTTP modules for handling non-native API integrations like this.

Step 5: Generate the Structured Summary with an LLM

The transcript is now text. This step passes it to a GPT-class model with a structured prompt that extracts exactly what HR needs: decisions made, action items with owners and deadlines, key discussion points, and any follow-up required.

1. Add an OpenAI module (or HTTP module calling https://api.openai.com/v1/chat/completions) after the transcription output.
2. Use gpt-4o as the model. GPT-3.5-turbo is cheaper but produces less consistent structured output for long transcripts.
3. Write a system prompt that specifies the output format. A proven structure for HR meetings:

You are an HR documentation specialist. Given the following meeting transcript, extract and return ONLY a JSON object with these keys:
- meeting_type: string (Interview / Performance Review / Disciplinary / Strategy / Other)
- participants_mentioned: array of strings
- key_decisions: array of strings
- action_items: array of objects with keys: owner, action, due_date_mentioned
- follow_up_required: boolean
- summary_paragraph: string (3-5 sentences, professional tone)
- compliance_flags: array of strings (any statements that may require HR legal review)

Do not include any text outside the JSON object.

4. Map the transcript text from Step 4 as the user message content.
5. Set max_tokens to 1500 for standard meetings. Increase to 2500 for strategy sessions with long transcripts.
6. Add a JSON Parse module after the LLM output to convert the response string into a structured data bundle that downstream modules can map field by field.

The compliance_flags field is the highest-value addition for HR teams. When the model flags a statement—accommodation requests, protected class references, disciplinary warnings—the summary routes to a separate review queue rather than auto-filing. This is not a legal review substitute; it is a first-pass filter that ensures sensitive content does not slip through silently.

Step 6: Route the Output and Handle Errors

Structured summaries now need to reach the right destination. This step handles both the happy path and the failure cases.

Happy Path Routing

1. Add an HRIS or ATS module to write the summary to the correct record. Map summary_paragraph, key_decisions, and action_items to the corresponding fields in your system.
2. If your HRIS does not have a native Make.com module, use an HTTP module with the HRIS REST API. Most modern HRIS platforms (BambooHR, Rippling, Workday, UKG) expose meeting note or document endpoints.
3. Add a parallel branch to send a Slack or Teams notification to the meeting organizer with the summary and action items. Use the action_items array to tag owners directly in the message.
4. Log every processed file to a Google Sheet or database with: filename, meeting type, processing timestamp, action item count, and whether compliance flags were raised. This creates an audit trail without additional effort.

Error Handling

Set up a dedicated error handler using Make.com’s built-in error routes. For a detailed build pattern, see how to set up routed error handling in Make with AI assistance.

• Transcription failure (API timeout, file too large): Route to a Slack alert asking a human to manually upload and process. Log the filename and error code.
• Malformed JSON from LLM: Add a retry with a simplified prompt. If the second attempt fails, route to the manual queue with the raw transcript attached.
• HRIS write failure: Store the summary in a holding Google Sheet and alert the system admin. Do not silently drop data.
• Compliance flag raised: Route to a separate Slack channel visible only to HR leadership and legal. Do not file to the general HRIS record until reviewed.

Step 7: Measure Results Against Your Baseline

Return to the numbers you logged in Step 1. After two weeks of live operation, measure:

• Time per meeting summary: Before vs. after. Most teams see manual summarization time drop from 20–45 minutes per meeting to zero active time.
• Action item tracking rate: Compare the percentage of meetings that produced a logged action item before and after automation.
• Summary filing consistency: Before automation, summaries end up in email threads, personal drives, or nowhere. After, 100% of processed meetings have a filed record.
• Error rate: Track how often the workflow requires manual intervention. A well-built workflow should require human fallback on fewer than 8% of recordings in the first month, dropping below 3% after prompt refinement.

Jeff’s rule applies here: 10 minutes of manual summarization per meeting per day equals one full work week lost per year per person doing it. For a team running 5 documented meetings daily, that is 5 weeks of recoverable capacity per year—before accounting for the quality improvement from consistent, structured output. For a fuller picture of what recovered time looks like at scale, see the case study on recovering $103K in annual labor hours with Make automation.

How to Know It Worked

The workflow is working when all five of these are true:

1. A meeting recording is saved to cloud storage and a structured summary appears in the HRIS within 15 minutes, with no human action taken.
2. Action items from the summary are visible in the team notification channel with owners tagged correctly.
3. The compliance flag queue has correctly routed at least one test recording that contained a flagged statement.
4. The audit log shows 100% of processed files with complete metadata entries.
5. The error handler has fired at least once during testing and produced an actionable alert—not a silent failure.

If any of these five are not true after your first week of live operation, stop adding meeting types and fix the gap first. Scaling a partially working workflow amplifies the failure mode, not the success.

Common Mistakes That Break This Workflow

• Using a personal account for cloud storage authentication. When that person leaves, the trigger breaks silently. Always use a service account or shared team account.
• Skipping the file extension filter. Without it, the scenario fires on every file in the folder—PDFs, slide decks, ZIP exports. The transcription API call fails and the error queue fills immediately.
• Using one prompt for all meeting types. An interview summary needs candidate evaluation structure. A disciplinary meeting summary needs documentation-grade precision. A strategy session summary needs decision capture. One generic prompt produces mediocre output for all three.
• Not parsing the LLM JSON output. Mapping raw LLM text directly to HRIS fields fails the moment the model returns a slightly different format. Always use a JSON Parse module between the LLM and the downstream write.
• Treating compliance flags as optional. The first time a sensitive conversation auto-files to a general HRIS record without review, the workflow becomes a liability rather than an asset. Build the compliance routing before you process real recordings.
• Going live without testing the error handler. Deliberately trigger each error condition during testing. An untested error handler is not an error handler—it is documentation that was never read.

For a broader view of what AI-assisted builds get wrong before they reach production, this list of 7 things an AI-built Make scenario gets wrong covers the failure patterns that appear most often.

Frequently Asked Questions

Does this workflow work for in-person meetings that are not recorded?

No. The trigger is a recorded audio or video file. For in-person meetings, the closest equivalent is a dedicated recording device that syncs to cloud storage, or a phone app like Otter.ai that auto-exports to a folder Make.com watches. The workflow itself is identical once the audio file exists in the watched folder.

How accurate is Whisper transcription for HR conversations?

Whisper achieves word error rates below 5% on clean audio in standard English. Accuracy drops on heavy accents, multiple overlapping speakers, and technical jargon specific to your industry. Review the first 10 transcripts manually and add a glossary or correction step to the prompt if specific terms are consistently misread.

Is it safe to send HR meeting recordings to OpenAI’s API?

OpenAI’s API does not use inputs from API calls to train models by default, which is materially different from the consumer ChatGPT product. However, your organization’s data classification policy and any applicable employment law governs what you can send to a third-party API. Disciplinary meetings, accommodation discussions, and medical leave conversations warrant legal review before going live. For teams with strict data residency requirements, Azure OpenAI Service provides the same models with contractual data residency controls.

What if our HRIS does not have a Make.com native module?

Use an HTTP module with your HRIS’s REST API. Most modern HRIS platforms expose a documents or notes endpoint. Feed the API documentation into the build process—see how to feed API docs into Claude to build Make HTTP modules for the exact approach. Legacy HRIS systems without REST APIs require a workaround: write to a structured Google Sheet that your HRIS can import via scheduled CSV sync.

Can this workflow handle multiple languages?

Whisper supports 57 languages and auto-detects language from audio. Set language to auto in the API call. The summarization prompt requires language-specific handling—either use a multilingual prompt or add a translation step between transcription and summarization. GPT-4o handles both tasks in a single call if you instruct it to translate and summarize simultaneously, which reduces API calls and latency.

Additional Reading

• What Is Automation-First? Why You Should Automate Before You Add AI
• 7 Questions to Ask Before You Automate Anything (The OpsMap Checklist)
• How to Run an OpsMap Audit Before Automating Anything
• What Is a Make Scenario? The Plain-English Guide for Zapier Users
• How to Build a Make Scenario With Claude: A Step-by-Step Walkthrough
• How to Set Up Routed Error Handling in Make With AI Assistance
• How to Feed API Docs Into Claude to Build Make HTTP Modules Without Native Connectors
• 7 Things an AI-Built Make Scenario Gets Wrong (And How to Catch Them)
• How to Evaluate a Make Scenario Built by AI Before It Goes to Production
• How One Ops Team Recovered $103K in Annual Labor Hours With Make Automation
• 10 Automations That Are Finally Easy to Build With Make + AI — No Developer Needed
• How a Non-Technical HR Team Started Building Their Own Automations With Make + AI
• 5 Automation Tasks AI Handles Well — and 5 It Still Gets Wrong
• AI-Assisted Make Builds vs. Manual Builds (2026): Which Is Better for Your Automation?
• 6 Ways the Make MCP Changes Automation Work for HR Teams