Key Takeaway
- 🎙️ Six copy-paste meeting minutes prompts turn any call recording or transcript into a decision document with named owners and real deadlines.
- 🧭 The workflow: record → transcribe → Prompt 1 (extract decisions) → Prompt 2 (assign owners) → Prompt 3 (flag risks) → minutes in under 10 minutes.
- 📋 Prompt 4 builds the follow-up tracker, Prompt 5 writes the client-ready summary, and Prompt 6 drafts the next meeting’s agenda from what is still open.
- 🔒 The safety rule: strip personal data before pasting transcripts into any AI tool, and always human-review names and commitments before sending.
- 👥 Works in any AI chat with any transcript language — Taglish calls included — and costs nothing but ten minutes after each meeting.
Every Filipino professional knows the meeting that ended with six people hearing six different conclusions. The client call ends, everyone nods, and two weeks later the deliverable is late because “I thought you were handling that.” The failure is not the meeting — it is the missing meeting minutes with names on decisions.
Handwritten notes capture what the note-taker could catch; memory captures nothing reliably; and the formal minutes template nobody has time for dies in the drawer.
The fix for meeting minutes is a system, not discipline: this week’s Prompt of the Day turns any meeting transcript into minutes with owners, deadlines, and follow-ups — in six prompts, in about ten minutes, with the phone recording you already have permission to make.
Table of Contents
Why Meeting Minutes Fail — and What Actually Fixes Them
The traditional minutes process fails for structural reasons, not personal ones. The note-taker is usually a participant — busy listening and contributing, writing partial sentences, and reconstructing them hours later from memory. Decisions get recorded as vibes (“team agreed to move fast”) instead of commitments (“Marco delivers the draft by Friday; Liza reviews Wednesday”).
And the minutes that do get written arrive days late, when the momentum they should have preserved has already scattered.
The AI version fixes each failure by design. The transcript is complete — nothing depends on typing speed. The extraction is verbatim — decisions read as spoken, not paraphrased into softness.
The owners and deadlines are forced fields — the prompt refuses to output a decision without a name attached, which is the entire accountability trick. And the whole system runs in ten minutes, same day, while memories are fresh.
The minutes stop being a chore and become the meeting’s actual product — and unlike every productivity promise of the past decade, this one requires nothing but a recording that already exists.
The Setup: One Recording, One Transcript, One Chat
Three pieces before the prompts. The recording: your meeting platform’s built-in recorder (Zoom, Teams, Google Meet all transcribe natively now) or a phone recorder for in-person sessions — with participants’ knowledge, always. The transcript: the platform’s auto-transcript, or a transcription tool of your choice; raw and unpolished is fine, the AI cleans it.
The chat: any capable AI assistant — the free tiers handle this workload easily. Total new hardware: zero. Total new subscriptions: none required.
One format note before the prompts: paste the transcript in sections if it is long — an hour-long call may exceed a paste window, and processing it in two halves with “continue” works cleanly. The prompts below assume the transcript is already in the conversation; each one builds on the previous output.
The Six Prompts, Copy-Paste Ready
Prompt 1 — Extract the decisions.
You are a precise meeting secretary. From the transcript below, list every DECISION made (things agreed to happen, directions chosen, approvals given). For each: the decision in one sentence, who proposed it, and any conditions mentioned. Quote the exact wording where a commitment is made. Ignore small talk and unresolved topics.Output as a numbered list.
Prompt 2 — Assign the owners.
For each decision above, identify the ASSIGNED OWNER — the person the transcript shows accepted responsibility. If a decision has no clear owner, list it under "UNOWNED — needs assignment" instead of guessing. Then list every explicit DEADLINE mentioned, attached to its decision.
Output as a table: Decision | Owner | Deadline | Source quote.
Prompt 3 — Surface the risks.
From the transcript, list every concern, objection, or risk that was raised but not resolved. For each: the issue in one sentence, who raised it, and what (if anything) was promised about it. Mark anything a participant promised to "check" or "follow up" as OPEN ITEMS.
Be exhaustive — these are the items meetings always forget.
Prompt 4 — Build the follow-up tracker.
Combine the decisions, owners, and unresolved items into a single follow-up tracker. Columns: Action | Owner | Due date | Status (New / In progress / Blocked). Sort by due date.
Add a "CHECK-IN DATE" column — propose a reasonable date to review each item based on its deadline and what the transcript suggests about urgency.
Prompt 5 — Write the client-ready summary.
Using the decisions and tracker, write a formal minutes summary suitable for sending to a client or senior stakeholder. Professional tone, no blame language, no internal disagreements quoted. Structure: Attendees · Key Decisions (with owners and dates) · Open Items and Next Steps · Next meeting date if mentioned.
Maximum 250 words.
Prompt 6 — Draft the next agenda.
From the open items in the tracker, draft the agenda for the next meeting of this group. Limit to 5 items maximum, ordered by urgency. For each: the item, the owner who will present it, and the outcome needed from the discussion (decision, update, or brainstorm).
End with a suggested meeting length.
Run them in order and the machine is complete: decisions extracted, owners named, risks surfaced, tracker built, summary drafted, next agenda queued — every artifact a meeting owes its participants, generated from the recording that already exists.
The Worked Example: a 15-Minute Client Call
Here is the machine on a realistic Filipino-freelancer call — a web-designer client check-in, transcript pasted raw from the platform.
Prompt 1 extracts five decisions, including one nobody noticed: the client mentioned — mid-small-talk — that "the logo probably needs a tweak before launch," and the AI correctly lists it as a decision with a condition ("budget pending").
Prompt 2's table assigns the designer to the mockup revision, the client contact to the brand-guide approval, and flags one decision UNOWNED — the hosting-renewal question that both sides assumed the other would handle. That single flag prevents the classic two-week silence.
Prompt 3 surfaces the risk the meeting smoothed over: the client's "we might have the content late" concern, unresolved, now visible in writing.
Prompt 4 turns everything into a tracker with check-in dates; Prompt 5 produces the 180-word client summary that makes the freelancer look operationally sharp; Prompt 6 drafts next week's agenda with the content deadline first. Total elapsed time from transcript paste to sent minutes: under twelve minutes, including the human review that Rule three requires.
The alternative — the memory-based workflow that lost the hosting question — costs a renewal lapse or an awkward client call. The comparison is the business case.
Variations for Different Meeting Types
The six prompts adapt with one-line modifications. For sales calls: add to Prompt 1 — "also extract every objection and every buying signal mentioned" — and the tracker becomes a deal-maintenance list. For project standups: Prompt 6 does the heavy lifting, and Prompts 1-2 compress into one (agile meetings produce few decisions, many blockers).
For client interviews: add to Prompt 3 — "extract every requirement or preference stated, even in passing" — because interviews hide requirements in small talk exactly the way the worked example's logo did.
For government-adjacent meetings — barangay, association, coop — Prompt 5's output becomes the secretary's draft minutes in the format the organization already expects.
The variation habit matters more than the variations themselves: after two weeks of use, most professionals customize one prompt permanently — a standing template that encodes their industry's vocabulary and their client's expectations. The pack is a starting point; the tailored version is the asset.
The Weekly Review Loop
The machine earns compounding returns when the trackers are reviewed on a schedule. The practice: Friday afternoons, fifteen minutes, open the week's trackers in sequence, update the Status column, and let Prompt 4's check-in dates drive the following week's calendar.
The professionals who adopt the loop report the same transformation: the feeling of "what did we decide?" disappears from their working life, replaced by a searchable record that answers in seconds. The meeting minutes stop being history and become infrastructure — the record the team trusts because it is complete, owned, and checked.
The Safety Rules
Rule one: consent and disclosure. Record and transcribe only with participants' knowledge — meeting platforms announce recording for a reason, and Philippine data-privacy law (RA 10173) governs personal data processing; a one-line disclosure at the meeting start covers the workflow.
Rule two: strip before you paste. For sensitive meetings, remove or mask personal identifiers the minutes do not need — the extraction prompts work fine on first names.
Rule three: human review before sending. AI-extracted minutes are drafts; the names, the commitments, and the deadlines must be human-checked before anything reaches a client — a misattributed action item costs more goodwill than the ten minutes saved. The system's value survives all three rules intact.
The Team Adoption Play
Introducing the machine to a team takes one meeting, not a memo.
Run it live: record the meeting as usual, run Prompts 1-2 visibly at the end, and share the extracted table on screen — the room sees its own decisions with names attached, and the objection ("AI will mishear us") dies against the accuracy of the demo — the meeting minutes produced live carry a credibility no after-the-fact summary ever does.
Assign the minutes role on rotation — ten minutes each, fair shares — and store the outputs in the shared drive, dated, searchable. Within a month the tracker becomes the team's operating memory: new members read the last three trackers and know the project's state.
For the OFW professional the adoption is even simpler: client calls across time zones are exactly where minutes matter most, because the follow-up happens across a time-zone gap where a missed detail waits a full day to surface.
The ten minutes of meeting minutes after the call is the cheapest project insurance available anywhere — and it renews daily at no cost.
Frequently Asked Questions
Do I need a paid AI subscription for this?
No — the six prompts run on free tiers of capable AI chats; long transcripts may need pasting in parts, but no subscription is required for the workflow.
What if my transcript is in Taglish?
Capable AI models handle Taglish directly — paste it raw; the extraction reads code-switched language well, and Prompt 5 can output the client summary in pure English regardless.
How accurate are the extracted decisions?
Verbatim transcripts extract cleanly, but sarcasm, overlapping speakers, and implicit agreements still need human review — which is why Rule three exists; treat the output as a first draft with the names double-checked.
Is recording meetings legal in the Philippines?
With participants' knowledge and consent, yes — the privacy law (RA 10173) requires lawful and disclosed processing of personal data; announce the recording and keep the minutes' distribution to participants.
Can this work for personal meetings too?
Yes — family decisions, condo-association meetings, and church-group planning all benefit; the prompts do not care who is in the room.
Where do I keep the outputs?
A dated file per meeting in the shared drive the team already uses — the tracker format makes the files searchable, and Prompt 6's agenda links each meeting to the last.
