A lightweight AI tool that captures short voice clips from real-world events (e.g., conferences, meetings), transcribes them using the Whisper API, and uses GPT to extract structured contact data and summaries. This information is then automatically logged into a Notion database and can trigger follow-up workflows.
Conferences, trade shows and networking breakfasts generate good conversations and very little usable data. Business cards get lost, notes are scrawled on the back of lanyards, and by the time a salesperson is back at a desk the details of who wanted what have blurred together. Contacts are entered days later, if at all, and follow-ups go out late and generic. For small sales teams without a dedicated sales operations person, the CRM ends up missing exactly the warm leads the event budget was meant to produce.
A voice note to CRM automation lets the salesperson do the one thing that is easy in a busy exhibition hall: talk into their phone for half a minute. Speech recognition turns the clip into text, an LLM pulls out the name, company, role, interest and next step, and a structured record lands in a Notion database. A follow-up email draft is prepared at the same time. The salesperson reviews the entry and the draft that evening instead of reconstructing the day from memory, and nothing depends on remembering to type details in later.
The salesperson records a short voice memo on their phone and shares it to a dedicated email address, cloud folder or form that the automation watches.
Make.com or Zapier sends the audio to OpenAI's Whisper speech-to-text API and receives a transcript, which is stored with the original recording for reference.
An LLM reads the transcript and returns structured fields: full name, company, role, contact details if mentioned, topics of interest, urgency and a suggested next step.
The automation checks Notion for an existing contact with the same name and company, then creates a new database entry or appends the note to the existing one.
A personalised follow-up email is drafted and held for approval; once the salesperson approves it, the workflow sends it and logs the date in Notion.
After meeting someone at a conference, a salesperson records a 30-second voice note: 'Met Jane Doe from Acme Inc, she's interested in our new widget.' The AI transcribes the note, creates a new contact 'Jane Doe' in a Notion CRM, and drafts a follow-up email, which is then sent via a Make.com automation.
OpenAI Whisper API, OpenAI API, Notion API, Make.com/Zapier, Mobile Voice Recorder
Voice notes about people you met are personal information, so the Australian Privacy Principles apply to how you collect, store and use them. Tell contacts you will follow up, and keep notes factual and professional because individuals can ask to access information held about them. Do not secretly record the conversation itself; state surveillance and listening devices laws restrict recording private conversations without consent, the rules vary between states, and a statutory tort for serious invasions of privacy has applied since June 2025. Follow-up emails are commercial electronic messages under the Spam Act 2003, so rely on consent such as an exchanged business card and interest in your offer, identify your business and include an unsubscribe option. Consider where OpenAI and Notion store data if notes contain sensitive details.
This solution requires careful evaluation of cost-benefit ratio. Consider implementation only after thorough analysis and risk assessment.
No. Notion suits small teams that already work there, but the same flow can write to HubSpot, Pipedrive, Airtable or another CRM with an API. The transcription and extraction steps stay the same; only the final action in Make.com or Zapier changes, along with the field mapping for your CRM's contact properties.
It depends mostly on how clearly the note is spoken and how noisy the environment is. Clear notes that follow a simple pattern are extracted reliably, while unusual names and email addresses spoken aloud are the most common errors. That is why the design keeps a quick human check before records are finalised and before emails go out.
Expect around two to four weeks. Most of the work is agreeing on the Notion database fields, writing and testing the extraction prompt, building duplicate checks and designing the approval step for emails. A short trial at a real event is the best way to tune prompts before the whole team relies on it.
Get expert guidance on implementing Voice-to-Notion Contact Logger for your business
Contact Our AI Experts