Last night, I Was Reworking a Quote at 11 p.m. and Realized the Recording Was Still on My Phone
On Tuesday evening, I finished a client meeting with 47 minutes of audio but only had time to jot down three lines. I still had to deliver a proposal and send a follow-up email the following week, so I stared at the chat history and stressed. For freelancers, the hardest part is often not the discussion itself, but turning that discussion into next steps. I have gotten stuck here too. Later, I filled in the gaps from memory and missed client commitments more than once.
Now I Use a Lightweight “Upload Recording → Transcribe → Generate Notes and Action Items” Workflow
The core is not complicated: first transcribe the audio, then ask AI to organize the text into client requirements, open questions, owners, and deadlines. My friend Lin Yan, a brand consultant in Hangzhou, uploads two interview recordings to her desktop every Wednesday morning. She generates a draft with the basic plan, then manually checks the figures and exact wording. Each call used to take her close to an hour to organize; now it takes about 15 minutes, and she no longer has to listen while taking notes. She does not send the output straight to clients. She treats it as a draft, personally verifying anything involving pricing, commitments, or judgment calls.
You Can Test It for Free—the Setup Is Easier Than Expected
Replication cost: Start with transcription and basic AI features for about RMB 0. Paid plans can begin at roughly RMB 30–100 per month. Setup takes about 30 minutes and requires no coding. On the first run, I simply created a project, clicked “Add Recording” or “Upload File,” and tested a five-minute recording. The only skills needed are copying, pasting, and checking the transcript. Knowing what a “prompt” is—a task instruction for AI—is enough. I would begin by asking it to organize only “requirements, action items, and open questions,” rather than building a complex system right away. If the work involves client privacy, I would get consent first and test with redacted text. This is not essential for everyone either. If I meet clients face to face and my handwritten notes are enough, there is no pressure to try it now.
Match the Workflow to Your Stage—There Is No Need to Build a Whole System
If I had just started and had only a few scattered consulting calls each week, I would use it to organize my own project reviews without rushing into a long-term subscription. With one or two steady clients, I would turn the process into a three-step template—recording to transcript, transcript to meeting notes, notes to follow-up email—and check for missing action items every week. While scaling, I would turn recurring client questions into templates, keep human review in place, and gradually track which types of notes save the most time. AI is not the final decision-maker. It is the assistant that lets me avoid transcribing everything by hand while preserving more of what happened in the room.