Claude Prompts for Zoom Meetings (Prep, Transcripts, Follow-Up)
Claude does not join your calls. It works on either side of them: describe the meeting and get the agenda and the script, then paste the Zoom transcript afterwards and get the decisions, the owners, and the recap email.
In short: This page contains 35 copy-paste ready prompts, organized into 5 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Hand-curated and tested by the AI Academy team.
Meeting Prep & Agendas
7 promptsAgenda Built Backwards From the Decision
1/35✨ What it does
Claude writes a time-boxed agenda built around the one decision the Zoom call must produce. You send it with the invite and cut any topic that does not serve that decision.
<context> Meeting: [title, length booked in Zoom, recurring or one-off] The decision or output this call must produce: [be specific] Attendees and what each brings: [names, roles] What people already know: [prior context] What usually goes wrong in this meeting: [overruns, no decision, one topic eats the hour] </context> <task> Build the agenda backwards from the required output. Time box each item against the booked length, cut anything that does not move the decision forward, and mark items that could be handled async instead of on the call. </task> <format> Agenda table: Minutes, Item, Owner, Purpose (decide, inform, discuss). Then a Move To Async list and a line stating what makes this meeting a success. </format>
Pro tip: Paste the agenda into the Zoom meeting description field so it rides along in the calendar invite. People read the invite; they rarely open an attached doc.
Pre-Read That Replaces the First Ten Minutes
2/35✨ What it does
Claude writes a pre-read that removes the usual set-the-scene section so the call can start at the actual discussion. You send it ahead of the meeting and skip the first ten minutes of recap.
<context> Background material: [PASTE THE DATA, EMAIL THREAD, OR DRAFT PROPOSAL] Meeting purpose: [what we are deciding] Attendees and their familiarity with the topic: [names, high or low context] Time I want to save on the call: [minutes] </context> <task> Write the pre-read that makes the usual set-the-scene section unnecessary. State the situation, the options, the trade-offs, and the specific question each attendee needs to arrive with an answer to. </task> <format> One page maximum: Situation in three sentences, Options with trade-offs, What We Need From You per attendee name, Questions to answer before the call. </format>
Pro tip: Send it 24 hours ahead and state in the invite that the call starts at the decision. Sending it an hour before guarantees nobody reads it.
Audit a Recurring Call That Nobody Wants
3/35✨ What it does
Claude compares what the recurring call actually delivers with what it claims, and makes an evidence-based case to shorten, thin, or kill it. You take that case to the owner and change the Zoom series if they agree.
<context> Recurring meeting: [name, frequency, length, headcount] Last three agendas or transcripts: [PASTE THEM] Why it exists officially: [stated purpose] What people say about it privately: [complaints] </context> <task> Work out what this meeting is actually delivering versus what it claims to. Identify which recurring items are status updates that could be written, which need live discussion, and what a shorter or less frequent version would lose. </task> <format> Three sections: What This Meeting Actually Produces, Items That Could Be Written Instead, Recommended New Shape with frequency, length and attendee list. Then the message announcing the change. </format>
Pro tip: Run this on the calendar cost. A weekly hour with eight attendees is over 400 person hours a year, and that number moves the decision faster than the complaints do.
Write the Invite Description People Actually Read
4/35✨ What it does
Claude writes the Zoom invite body: the purpose, the prep, whether it is recorded, and who can safely decline. You paste it into the invite and drop people who now have cover to skip.
<context> Meeting: [title, length] Why each person is invited: [name, reason] What attendees must do before joining: [read, prepare numbers, decide internally, nothing] Whether the call will be recorded: [yes or no] Agenda: [paste it if you have one] </context> <task> Write the description that goes in the Zoom invite body. Lead with the purpose in one sentence, list the prep required, and be explicit about whether attendance is required or optional so people can decline without checking first. </task> <format> Invite body under 120 words: Purpose line, Prep Before Joining, Agenda in three or four lines, Required vs Optional attendees, recording notice if recording. </format>
Pro tip: Say who is optional explicitly. Most people accept every invite by default because declining without permission feels risky, and the call gets bloated.
Triage the Attendee List
5/35✨ What it does
Claude splits the invite list into needed for the decision, needed for one topic, and optional, plus cover for the people you drop. You send the smaller invite and a courtesy note to the ones left off.
<context> Meeting purpose and required output: [what] Current invite list: [names and roles] Who is genuinely needed to reach the decision: [your instinct] Politics involved: [who will be offended if left out] </context> <task> Split this list into three groups: needed to make the decision, needed only for one specific agenda item, and recap only. For the second group, say which item they should join for. For the third, say what they get instead of an invite. </task> <format> Three labelled lists with one line of reasoning per name. Then a short note I can send to anyone I am dropping, phrased so it reads as respecting their time. </format>
Pro tip: For the item-specific group, give them the time window in the invite. Zoom shows join time, so people can drop in for their fifteen minutes and leave.
Prep Brief Before an External Call
6/35✨ What it does
Claude builds a one-page brief on the people, the history, and likely objections before the external call. You read it once, then walk into Zoom with your goal for [what I want to leave with] in mind.
<context> Who I am meeting: [company, names, titles] Public information I gathered: [PASTE THEIR WEBSITE COPY, RECENT NEWS, LINKEDIN SUMMARIES] Our history with them: [paste email thread or account notes] My goal for this call: [what I want to leave with] </context> <task> Build the prep brief. Summarize where the relationship stands, infer what each person on their side likely cares about from their role, list the questions I should ask, and predict the three objections most likely to come up with an answer for each. </task> <format> Brief with: Relationship status, Person by person read, My five questions, Likely objections with responses, What a good outcome looks like. </format>
Pro tip: Keep the objections section open in a second window during the call. Reading a prepared answer is a lot easier than inventing one while sharing your screen.
Run-of-Show for a Multi-Presenter Call
7/35✨ What it does
Claude builds a minute-by-minute run-of-show with clock times, screen-share handovers, and host cues. You share it with presenters and run the session from those clock times, not from requested durations.
<context> Session: [title, total length, audience size] Presenters and their segments: [name, topic, requested minutes] Who screen shares when: [list] Interactive parts: [polls, breakout rooms, live Q&A] Host and co-host: [names] </context> <task> Build the run-of-show. Assign clock times not durations, note every handover including who takes screen share control, place the interactive moments where attention normally dips, and add the buffer that these sessions always need. </task> <format> Table: Clock time, Segment, Who leads, Screen share owner, Host action. Then a Contingency list covering an overrunning presenter and a presenter who does not show. </format>
Pro tip: Give the co-host the run-of-show too. Managing the recording, the polls and the room while presenting is what causes the visible fumbles.
XML tags are just the start. Learn the full Claude workflow.
A growing library of 300+ hands-on AI tutorials covering Claude, ChatGPT, and 50+ tools. New tutorials added every week.
Transcript Summaries (Paste the Transcript)
7 promptsClean Up a Raw Zoom Transcript First
8/35✨ What it does
Claude cleans the raw Zoom VTT into a readable transcript, with your speaker names and corrected terms. You archive that version and use it for every later summary.
<context> Raw transcript downloaded from the Zoom recording: [PASTE THE VTT OR TXT CONTENT] Real names for the speaker labels: [Speaker 1 is X, or the display names shown] Terms the transcription got wrong: [product names, acronyms, people's names] </context> <task> Clean this into a readable transcript. Strip the timestamp and cue number lines, merge consecutive lines from the same speaker into paragraphs, fix the misheard terms I listed, and mark anything unintelligible rather than guessing at it. </task> <format> Clean transcript with speaker name followed by their paragraph. Keep a timestamp only at each topic change. End with a list of terms you corrected so I can check them. </format>
Pro tip: Zoom cloud recordings produce a .vtt file you download from the recording page. Do this cleanup pass first and every later summary gets noticeably more accurate.
Decisions and the Reasoning Behind Them
9/35✨ What it does
Claude extracts every decision the call actually made, with the reasoning, rejected alternatives, and who made the call. You paste the log into the project notes and treat preferences as still open.
<context> Meeting transcript: [PASTE TRANSCRIPT] Meeting purpose: [what it was for] Who has decision authority: [name and role] </context> <task> Extract every decision the call actually made. For each one, capture the reasoning given, the alternatives that were rejected and why, and who made the call. Separate real decisions from things people agreed sounded good but never confirmed. </task> <format> One block per decision: Decision, Made by, Reasoning, Rejected alternatives, Confidence that this was truly decided. Then a list headed Discussed But Not Decided. </format>
Pro tip: The rejected alternatives are what stops the same debate reopening in six weeks. Store them with the decision, not in a separate note.
What Each Person Committed To, By Name
10/35✨ What it does
Claude goes speaker by speaker and pulls everything each person said they would do, with the quote and a firm or vague rating. You send the firm lines to those people and clarify the vague ones.
<context> Transcript with speaker labels: [PASTE TRANSCRIPT] Attendees and roles: [names] Date of the call: [date] </context> <task> Go speaker by speaker and pull out everything each person said they would do, including soft commitments phrased as I can take a look at that. Rate each one as firm, vague, or someone else volunteering them, and quote the line it came from. </task> <format> Section per person: Commitment, Firmness, Date if stated, Quote. End with a Nobody Owns This list for work everyone agreed was needed with no name attached. </format>
Pro tip: Zoom labels speakers by their display name, so someone joining from a shared room device shows up as a room name. Map those to real people in the context block first.
Ten-Line Summary for People Who Skipped
11/35✨ What it does
Claude compresses the hour into ten lines: decisions, new information, and work now waiting on someone. You send that version to [colleagues who declined, a wider channel, leadership] and skip the full recording.
<context> Transcript: [PASTE TRANSCRIPT] Meeting: [what it was, how long it ran] Audience for the summary: [colleagues who declined, a wider channel, leadership] </context> <task> Compress this into ten lines maximum. Include only what changes what someone does next: decisions, new information, and anything now waiting on a person outside the call. Cut the discussion, the tangents, and the parts that ended where they started. </task> <format> Ten single-line bullets, each starting with the subject. No preamble, no closing line, no restating the meeting title. </format>
Pro tip: Ask for the same thing twice at ten lines and at three lines. The three-line version is what you paste into a channel; the ten-line one goes in the doc.
Pull Out Only the Part About My Topic
12/35✨ What it does
Claude finds every moment [subject, project, or account] came up, including passing mentions, and leaves the rest out. You read only that extract and ignore the rest of the call.
<context> Full transcript of a long call: [PASTE TRANSCRIPT] The only topic I care about: [subject, project, or account] Why I care: [I own it, I am affected by it, I have to report on it] </context> <task> Find every moment this topic came up, including passing mentions and references made without naming it directly. Summarize what was said about it, who said it, and whether anything about it was decided. Ignore the rest of the call entirely. </task> <format> Chronological list of mentions with speaker and approximate timestamp, then a short synthesis of what this means for my topic, then anything it now requires from me. </format>
Pro tip: Ask for approximate timestamps in the output. You can jump straight to that point in the Zoom recording to hear the tone, which the transcript loses.
Questions Raised and Never Answered
13/35✨ What it does
Claude lists every question asked and flags the ones that were deflected, answered sideways, or drowned out. You put those unanswered questions on the next agenda.
<context> Transcript: [PASTE TRANSCRIPT] Meeting purpose: [what] Who was supposed to have the answers: [names and their areas] </context> <task> Find every question asked in this call and check whether it actually got an answer. Flag the ones that were deflected, answered with something adjacent, or lost when the conversation moved on. Note who asked and who should have answered. </task> <format> Table: Question, Asked by, Answered fully or partially or not at all, Who owes the answer. Then rank the unanswered ones by how much they matter to the decision. </format>
Pro tip: Unanswered questions from a call make the strongest agenda for the next one. They are already specific and someone already cared enough to ask.
Merge the Transcript, the Chat and Your Notes
14/35✨ What it does
Claude builds one meeting record from the transcript, the chat export, and your notes, folding chat in at the right moment. You save that record and use your notes only where Claude flagged a conflict.
<context> Zoom transcript: [PASTE TRANSCRIPT] In-meeting chat export: [PASTE THE CHAT TXT] My own notes typed during the call: [PASTE NOTES] </context> <task> Build one record from all three sources. Fold the chat messages in at the point of the conversation they relate to, use my notes to resolve anything the transcript garbled, and flag any place where my notes and the transcript disagree instead of silently picking one. </task> <format> Single merged record organized by topic, with a source tag on anything that came only from chat or only from my notes, plus a Conflicts section listing disagreements between sources. </format>
Pro tip: Zoom saves the in-meeting chat as a separate text file alongside a cloud recording when that setting is on. The links and side comments in it never appear in the transcript.
Action Items & Follow-Ups
7 promptsAction List With Owners and Realistic Dates
15/35✨ What it does
Claude extracts every action item with an owner and a date, marking proposed dates when none was stated. You confirm the proposed dates with owners, then paste the list into the follow-up.
<context> Transcript: [PASTE TRANSCRIPT] Date of the call: [date] Attendees and their current workload: [names, rough capacity] Hard deadlines that already exist: [dates] </context> <task> Extract every action item. Where a date was stated, use it. Where none was given, propose one and mark it as proposed rather than agreed. Where no owner was named, assign the most plausible person and flag it clearly as an assumption for me to confirm. </task> <format> Table: Action, Owner, Due, Stated or proposed, Blocked by. Sort by due date. Then a short list of the items you had to guess an owner for. </format>
Pro tip: Never send an action list with silently invented owners. Marking the guesses gets them corrected in the reply instead of quietly ignored.
Personal Chase Message for Each Owner
16/35✨ What it does
Claude writes one short chase message per owner covering only their items. You send each message in [Slack DM, email, the meeting thread] so nobody has to hunt their line in a group list.
<context> Action list from the call: [PASTE THE ACTIONS WITH OWNERS AND DATES] Relationship to each owner: [peer, report, another team, client] How long since the call: [days] Where I will send these: [Slack DM, email, the meeting thread] </context> <task> Write one short message per owner covering only their items. Include the context they need to start without rereading the transcript, and end each with a question they can answer in one line. </task> <format> One block per person, each a complete message under 60 words, headed with the person's name and the channel I should send it in. </format>
Pro tip: Individual messages get answered several times more often than a group post with a table. Nobody feels personally addressed by a shared action list.
Turn Actions Into Ticket Descriptions
17/35✨ What it does
Claude rewrites each action as a ticket with context and acceptance criteria for [Jira, Linear, Asana, and the fields you require]. You create the tickets and edit titles to match team convention.
<context> Actions from the call: [PASTE THE ACTION LIST OR THE RELEVANT TRANSCRIPT SECTION] Tracker and its fields: [Jira, Linear, Asana, and the fields you require] Project or board: [name] Team conventions: [how you write titles, what a good description contains] </context> <task> Turn each action into a ticket a colleague could pick up without having attended. Add the context from the discussion, state the acceptance criteria, and note the open questions that need answering before work starts. </task> <format> One ticket per action: Title, Description, Acceptance criteria, Open questions, Owner, Estimate if inferable. Then the same set as comma separated values matching my tracker fields. </format>
Pro tip: Ask for the CSV version with your exact column headers. Most trackers accept a CSV import, which beats creating fifteen tickets by hand.
Build the Next Agenda From What Was Left Open
18/35✨ What it does
Claude drafts the next agenda from what this call left unfinished, prioritizing blockers and who must be in the room. You send that agenda with the next Zoom invite.
<context> Transcript of the call that just ended: [PASTE TRANSCRIPT] Next meeting: [date, length, same attendees or not] Standing items we always cover: [list] </context> <task> Draft the next agenda from what this call left unfinished. Prioritize items that are blocking someone, note which need a specific person present, and drop anything that was resolved in the discussion even if it was on the previous agenda. </task> <format> Agenda with time boxes: Item, Why it is here, Who must attend for it, Minutes. Then a Carried Over Twice list flagging anything that has now been deferred more than once. </format>
Pro tip: The carried-over-twice list is a signal, not an agenda item. Anything deferred twice needs a decision outside the meeting or it will be deferred a third time.
Find the Work Nobody Agreed to Own
19/35✨ What it does
Claude finds the work everyone agreed was necessary but nobody picked up, including phrases like we should probably. You assign each item to the suggested owner or drop it on purpose.
<context> Transcript: [PASTE TRANSCRIPT] Team and what each person owns: [names, scope] Who runs this project: [name] </context> <task> Find everything the group agreed was necessary but never assigned. Include work hidden inside phrases like we should probably and someone needs to. For each item, suggest the most logical owner based on what people own, and say what makes it ambiguous. </task> <format> Table: Unowned work, Quote it came from, Suggested owner, Why ownership is unclear. Then a short message I can send to propose the assignments. </format>
Pro tip: Send the proposed assignments as a message rather than adding names to a doc. Assigning someone silently in a spreadsheet gets you a surprise, not a commitment.
Risk and Blocker Register From the Call
20/35✨ What it does
Claude builds a risk register from concerns people raised and then moved past, marking which are new. You add the new ones to the project tracker and estimate them before the next call.
<context> Transcript: [PASTE TRANSCRIPT] Project and deadline: [name, date] Known risks already tracked: [list] </context> <task> Pull out every risk, dependency and blocker mentioned, including the ones raised in passing and dropped. Mark which are already on our register and which are new. For each new one, estimate likelihood and impact from what was said, and note who would need to act. </task> <format> Register table: Risk, New or known, Likelihood, Impact, Owner, Mitigation discussed or none. Then the two or three that deserve escalation this week. </format>
Pro tip: Risks usually surface as a throwaway line rather than an agenda item. A transcript pass catches the ones that never made it into anyone notes.
Check This Call Against Last Call Commitments
21/35✨ What it does
Claude compares last call's actions to this week's transcript and marks each as done, in progress, silently dropped, or reassigned. You bring the dropped ones back by name in the next meeting.
<context> Previous meeting actions: [PASTE THE PREVIOUS ACTION LIST] Transcript of this week's call: [PASTE TRANSCRIPT] Time between the two: [days or weeks] </context> <task> Compare the two. Mark each previous action as done, in progress, silently dropped, or reassigned, using what was said in the new transcript as evidence. Highlight anything that was promised again with a new date. </task> <format> Table: Previous action, Owner, Status, Evidence from this call. Then a short section headed Promised Twice listing items now on their second or third deadline. </format>
Pro tip: Run this every two or three cycles rather than weekly. The pattern of repeat slippage is only visible across several meetings, and it is the useful signal.
These prompts give you the what. Tutorials give you the why.
Learn when to use extended thinking, how to build Claude Projects, and workflows that compound. 300+ tutorials and growing.
Recap Emails After the Call
7 promptsInternal Stakeholder Recap Email
22/35✨ What it does
Claude writes a scannable internal recap that puts required actions above the detail. You send it to [names, roles, interests] and BCC people who only need awareness.
<context> Transcript or my notes: [PASTE] Recipients and what each cares about: [names, roles, interests] Their involvement: [attended, invited but absent, not invited but affected] What I need them to do after reading: [nothing, approve something, adjust a plan] </context> <task> Write the recap email. Lead with what changed rather than what was discussed, keep it scannable for someone reading on a phone, and put anything requiring their action in a clearly separated block near the top. </task> <format> Email with a subject line under 60 characters, a two-sentence opening, Decisions, Action Required From You with names, Everything Else in three bullets. Under 250 words. </format>
Pro tip: Write the subject line as the outcome, for example: pricing approved, launch moves to March 3. Half your recipients will act on the subject line alone.
Client Recap That Confirms What Was Agreed
23/35✨ What it does
Claude writes a client confirmation email that restates what was agreed and keeps what was only floated clearly open. You send it after you remove anything in [what I should not put in writing yet].
<context> Call transcript or notes: [PASTE] Client and attendees: [company, names] What was agreed: [scope, price, dates, conditions] What is still open: [items] Commercial sensitivity: [what I should not put in writing yet] </context> <task> Write the confirmation email. Restate what was agreed precisely enough that it stands as a record, separate agreed items from things still under discussion so neither side assumes too much, and invite corrections with a deadline. </task> <format> Email with: opening thanks line, Agreed On Today as a numbered list, Still Open as a list with owners, Next Steps with dates, a closing line asking them to reply with corrections by a stated date. </format>
Pro tip: The reply-with-corrections deadline is what turns a recap into a record. Silence against a stated deadline is far more useful than silence against nothing.
Recap for the Executive Who Dropped Off Early
24/35✨ What it does
Claude writes a short catch-up covering only what happened after they left, plus anything they still need to approve. You send it to that executive and skip the part of the call they already heard.
<context> Transcript: [PASTE TRANSCRIPT] When they left: [approximate time or topic] Their role and what they care about: [title, priorities] What happened after they left: [key points] Anything they need to approve: [items] </context> <task> Write a short message covering only what happened after they dropped, plus anything from before that was later reversed. Flag anything that assumes their approval, and ask for the one decision you need from them. </task> <format> Message under 120 words: what changed after you left, what now needs your approval, one direct question. No agenda replay, no thanks-for-joining opening. </format>
Pro tip: Say explicitly if a decision was made in their absence that they may want to revisit. Discovering that later is what turns a small decision into a reversal.
Discovery Call Recap for the CRM
25/35✨ What it does
Claude turns the discovery call into CRM-ready notes, filling each [MEDDIC, BANT, or your own fields] field from quoted evidence. You paste the fields into the CRM and leave blank anything Claude could not quote.
<context> Discovery call transcript: [PASTE TRANSCRIPT] Prospect: [company, size, industry] Attendees on their side: [names, titles] Our qualification framework: [MEDDIC, BANT, or your own fields] CRM fields I have to fill: [list] </context> <task> Turn this call into CRM-ready notes. Fill each qualification field from what was actually said, quote the evidence, and leave fields empty with a note rather than inferring. Separate what the prospect stated from what I am assuming. </task> <format> Field by field: value, evidence quote, confidence. Then Next Steps with dates, Red Flags, and the three questions to ask on the next call to fill the gaps. </format>
Pro tip: Insisting on a quote per field kills the usual problem, which is a confident CRM entry based on something the prospect never actually said.
Recap Email With Proposal Next Steps
26/35✨ What it does
Claude writes a sales follow-up that mirrors the buyer language and proposes a dated next step toward a proposal. You send it the same day and put the proposed date on the calendar.
<context> Sales call transcript: [PASTE TRANSCRIPT] What they said they need: [requirements in their words] Objections raised: [list] Our offer and pricing: [details] Decision process they described: [who signs, timeline] </context> <task> Write the follow-up email that moves this to a proposal. Mirror their language rather than our product vocabulary, address the objection that will decide this deal, and propose one specific next step with a date instead of asking how they want to proceed. </task> <format> Email under 200 words: what I heard, how we address the main objection, what happens next with a date and a named owner on each side. Then a one-line alternative next step if the first is too fast for them. </format>
Pro tip: Have Claude list the exact phrases the prospect used for their problem and reuse those verbatim. Translating their words into your feature names is what makes follow-ups feel generic.
Recap When Nothing Was Actually Decided
27/35✨ What it does
Claude writes an honest recap that says plainly nothing was decided, captures each position, and sets up the next attempt. You send it while the stall is fresh so the next call starts further along.
<context> Transcript of a call that ended without a decision: [PASTE TRANSCRIPT] What it was supposed to decide: [the question] Why it stalled: [missing data, missing person, disagreement, ran out of time] Deadline pressure: [what happens if this keeps slipping] </context> <task> Write the honest recap. Say plainly that no decision was reached, capture the positions accurately so the next session does not restart from zero, and specify exactly what has to happen before the next attempt, with owners. </task> <format> Message with: No decision reached, Where each position stands, What is blocking a decision, What we need before next time with owners and dates, proposed date to reconvene. </format>
Pro tip: Writing that no decision was reached is what forces the next call to be scheduled. A vague productive discussion recap lets the topic drift for weeks.
Recap Email in the Client Language
28/35✨ What it does
Claude writes the recap in the recipient's working language, with an English summary of what you sent, keeping required terms in English. You send the foreign-language version and keep the English summary for your file.
<context> Transcript (in the language it was spoken): [PASTE TRANSCRIPT] Recipient and their working language: [name, language] Formality expected: [formal, standard business, informal] Terms that must stay in English: [product names, contract terms, legal wording] </context> <task> Write the recap in their language at the right level of formality for their business culture. Keep the listed terms untranslated, and flag anything from the call where a translation could change the commercial meaning. </task> <format> Full email in the target language, then a short English summary of what it says, then a list of any wording you flagged as commercially ambiguous. </format>
Pro tip: Always request the English summary alongside. Sending a message in a language you cannot fully verify is the fastest way to agree to something you did not intend.
Difficult Conversations & Facilitation
7 promptsScript a Performance Conversation on Video
29/35✨ What it does
Claude writes a branching script for the performance conversation on video, including how to open and where to stop talking. You use the opening, then follow the branch that matches how they react.
<context> Person and their role: [name, tenure, role] The specific behaviour or results gap: [factual examples with dates] What I have already raised and when: [prior conversations] Outcome I need: [change in behaviour, plan, formal process] How they usually react: [defensive, shuts down, agrees then nothing changes] </context> <task> Script the conversation for a video call. Open in a way that does not trigger a defensive shutdown in the first thirty seconds, present the evidence factually, and build in the pause where I stop talking and let them respond. Prepare my response to each of their likely reactions. </task> <format> Script with: Opening lines verbatim, Evidence to state, The question that hands it to them, Branch responses for each likely reaction, Closing agreement with a date. Add a note on what to say if they get emotional. </format>
Pro tip: On video you cannot read body language well, so the pause feels longer than it is. Rehearse the silence after your question and do not fill it.
Deliver Bad News to a Client on the Call
30/35✨ What it does
Claude prepares a structure for telling the client the bad news, with honest responsibility, impact, and answers to the hard questions. You say it on the call in that order, then send the written follow-up.
<context> The bad news: [delay, budget overrun, missed commitment, defect] Our responsibility in it: [honest assessment] Impact on them: [concrete] What we are doing about it: [remediation, timeline] What I can offer: [credit, scope change, escalation contact] Relationship history: [strong, strained, new] </context> <task> Prepare the call. Lead with the news rather than building up to it, be specific about what we got wrong without over-apologizing, and move quickly to the remediation. Prepare answers for their hardest questions, including whether this will happen again and whether they should be worried about the rest of the work. </task> <format> Opening statement verbatim, then Facts to state, Remediation, Hard questions with answers, Concessions in order of what to offer first. Add what not to say. </format>
Pro tip: Do this on video rather than email even though email feels safer. Bad news read in writing has no tone, and clients fill that gap with the worst interpretation.
Facilitation Plan for a Call One Voice Dominates
31/35✨ What it does
Claude gives you facilitation structures and exact phrases to rebalance who talks when one voice dominates. You use those phrases live and protect time for the quiet people on their topic.
<context> Meeting and purpose: [what] Attendees: [names, roles, seniority] Who dominates and how: [talks over people, long monologues, restates their point] Who goes quiet and why: [names, seniority gap, language, remote disadvantage] What I need from the quiet people: [their expertise on which topic] </context> <task> Design the facilitation. Use structures that make airtime a rule rather than a request, give me the exact phrases to interrupt a monologue without humiliating anyone, and give me the direct invitations that get a quiet expert to speak. </task> <format> Plan with: Structural choices for the agenda, Interrupt phrases word for word, Invitation phrases per quiet attendee, and one line on what to do if the dominant person ignores the structure. </format>
Pro tip: Round-robin works far better on video than in a room, because the host controls the order and there is no crosstalk to fight through.
Break a Deadlock in the Second Half of the Call
32/35✨ What it does
Claude gives you a time-boxed plan for the remaining minutes that reframes the stuck question into a decision. You run the second half from that plan and end with a recorded outcome, even a small one.
<context> What the deadlock is about: [the question] Positions and who holds them: [names, positions] What has already been tried: [arguments, data, escalation] Time left in the call: [minutes] What happens if we leave with nothing: [consequence] </context> <task> Give me a plan for the remaining time. Reframe the question so it is not the same argument again, identify what evidence would actually settle it, and prepare a fallback that produces a decision even if agreement is impossible. </task> <format> Plan with: Reframing question to ask out loud, Two structures to try in order with timings, What would resolve it factually, Fallback decision path, and the exact words to close the call either way. </format>
Pro tip: The fallback matters most. Deciding who decides, or setting a date to decide with named evidence required, still counts as leaving with an outcome.
Neutral Script for a Cross-Team Conflict Call
33/35✨ What it does
Claude writes a neutral facilitation script for the call between the two stuck teams, including what is not negotiable. You run the script as [neutral facilitator, one of the parties, leadership] and keep your own side out of the framing if you are meant to be neutral.
<context> Teams involved and the conflict: [who, what it is about] Each side's position and what is underneath it: [stated, actual] History: [how long, what has been tried] My role: [neutral facilitator, one of the parties, leadership] Constraint: [what is genuinely not negotiable] </context> <task> Write the facilitation script. Open by stating the shared goal and the ground rules, structure it so each side is heard before any solving starts, and keep the language neutral so neither side hears you taking the other one's framing. </task> <format> Script with: Opening and ground rules verbatim, Phase timings, Neutral phrasings for the contested terms, Questions that move from positions to interests, Closing agreement structure with owners. </format>
Pro tip: Watch for loaded vocabulary. Ask Claude to list contested terms with a neutral replacement, because using one side words signals which side you are on.
Design Breakout Rooms for a Working Session
34/35✨ What it does
Claude designs breakout room composition, briefs, and a report-back format that fits [minutes]. You set the Zoom rooms from that plan and hold the report-back to the clock.
<context> Session: [purpose, total length, number of attendees] Output required from the breakouts: [decisions, ideas, drafts] Participants: [names, functions, seniority mix] Time available for breakouts: [minutes] Experience level of the group with this format: [used to it or not] </context> <task> Design the breakout structure. Decide group size and composition and say why, write the brief each room gets, set the time so groups do not stall or run dry, and design the report-back so it does not eat the rest of the session. </task> <format> Plan with: Room composition and reasoning, Brief to paste into the room, Timing including the warning at the halfway point, Report-back format with a per-room time limit, and what the host does during breakouts. </format>
Pro tip: Broadcast a message to all rooms at the halfway mark. Groups without a time signal spend the first ten minutes on introductions and the last two panicking.
Debrief Your Own Performance From the Transcript
35/35✨ What it does
Claude critiques your own contributions line by line: where you talked past someone, answered the wrong question, or lost the goal. You pick one habit to change and watch for it on the next call.
<context> Transcript of a call that did not go how I wanted: [PASTE TRANSCRIPT] My name in the transcript: [display name] What I was trying to achieve: [goal] What actually happened: [outcome] </context> <task> Review my contributions specifically. Show me where I talked past someone, where I answered a question that was not asked, where I conceded too early, and where I missed an opening someone gave me. Be direct rather than encouraging, and quote the lines. </task> <format> List of specific moments: quote, what I did, what the better move was and why. Then one paragraph on the pattern across all of them, and two things to change in the next call. </format>
Pro tip: Ask for directness explicitly and say you do not want encouragement. Without that instruction the feedback comes back softened to the point of being useless.
Free tool
Transcript Summarizer
Turn a meeting or video transcript into a summary with action items.
Frequently Asked Questions
Prompts are the starting line. Tutorials are the finish.
A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.
7-day free trial. Cancel anytime.
Related guides