AI Call Summary Templates: 6 Formats, Examples and Prompts

Updated 18 October 2026. New guide: six copy-paste call summary templates with fictional examples, transcript prompts, and the built-in summary options in Zoom, Google Meet and Microsoft Teams.

An AI call summary is a short, structured record of a call, written by a language model from the transcript. It says who attended, why the call happened, what was decided, who will do what by when, and what is still open. The best summaries follow a fixed template per call type, use only facts that were said, and get a quick human check before they reach your CRM.

Below are six templates to copy: sales discovery, support, client kickoff, internal meeting, recruiting screen and account review. Each has a short filled-in example; all names, companies and numbers in them are fictional. Then come the prompts and the ways to automate the job.

Related: Amazon Product Description Generator: AI Prompts and 2026 Rules

Quick reference: which call summary template to use

Call type	Must capture	Where it goes
Sales discovery call	Pain, current tool, budget, decision process, timeline, next step	CRM deal or contact
Support call	Problem, steps tried, fix or workaround, status	Help desk ticket
Client kickoff	Goal, scope, owners, milestones, risks	Project tool and client email
Internal meeting	Decisions, action items, blockers	Team channel or wiki
Recruiting screen	Evidence per criterion, logistics, recommendation	Applicant tracking system
Account review	Results, health, risks, renewal signals	CRM account record

What a good AI call summary contains

Whatever the call type, a useful summary answers the same questions in the same order:

1. Header: date, call type, attendees with roles, related account, deal or ticket.
2. Purpose: one sentence on why the call happened.
3. Key points: three to six bullets that will matter later.
4. Decisions: what was agreed, in plain words.
5. Action items: task, owner and due date.
6. Open questions and risks: what is unresolved.
7. Next step: the next meeting or milestone, with a date.

Two rules keep summaries trustworthy. Use only what was said, and write "Not discussed" instead of guessing. Copy numbers, dates, prices and commitments exactly, because people act on them.

6 call summary templates (copy and paste)

Each template is plain text, so it works in an email, a CRM note or a prompt. Delete fields you never use.

1. Sales discovery call summary template

Discovery call summary
Date: | Account: | Deal:
Attendees (name, role):
Why they took the call:
Pain points (their words):
Current solution and what it costs them:
Budget: (amount or "Not discussed")
Decision process and people involved:
Timeline and trigger:
Objections or concerns:
Next step (what, who, date):
Action items (task, owner, due):
Open questions:

Example (fictional): Northwind Dental. Office manager Priya Shah wants online booking: front-desk staff spend about 10 hours a week booking by phone. Current tool: desktop scheduler, no integrations. Budget: not discussed. The owner signs off. Goal: live before summer. Concern: patient data privacy. Next step: demo with the owner, 19 March. Action: rep sends a security overview by 14 March.

2. Support call summary template

Support call summary
Date: | Customer: | Ticket:
Product and version:
Problem (customer's words):
Impact (who is blocked, since when):
Steps already tried:
Cause (if found):
Fix or workaround given:
Status: resolved / escalated / waiting on customer
Follow-up (what, who, date):

Example (fictional): Bluefin Logistics, ticket 4821. Invoices export as blank PDFs since Monday's update, blocking month-end billing. The customer cleared the cache and tried two browsers. Cause: the update reset a template setting. Workaround: reselect the invoice template in Settings. Status: resolved. Follow-up: agent checks Friday's export.

3. Client kickoff meeting summary template

Kickoff summary
Date: | Client: | Project:
Attendees and roles (client side, our side):
Business goal (one sentence):
Success measures:
In scope:
Out of scope:
Milestones and dates:
Owners (decisions, content, approvals):
Risks and dependencies:
Communication (channel, meeting rhythm):
Action items (task, owner, due):

Example (fictional): Harbor Bakery website rebuild. Goal: take catering orders online. Success: 30 online catering orders a month by month three. In scope: site, order form, card payments. Out of scope: loyalty app. Milestones: design sign-off 2 May, launch 6 June. Approvals: Maria. Risk: menu photos not ready. Action: Maria sends menu and prices by 25 April.

If the kickoff leads to a build, this summary feeds a software requirements document.

4. Internal meeting summary template

Meeting summary
Date: | Team: | Meeting:
Attendees:
Purpose:
Decisions made:
Action items (task, owner, due):
Blockers and who can remove them:
Parked topics:

Example (fictional): Weekly operations meeting. Decision: the courier cut-off moves from 3 pm to 2 pm next Monday. Actions: Sam updates the checkout notice by Friday; Lee emails wholesale clients by Thursday. Blocker: label printer driver (Sam calls the vendor). Parked: holiday staffing.

5. Recruiting screen summary template

Screening call summary
Date: | Candidate: | Role: | Interviewer:
Evidence for each must-have criterion:
Gaps or concerns (job-related only):
Candidate's questions:
Logistics (notice period, location, salary range):
Recommendation: advance / hold / decline, with reason
Next step:

Example (fictional): Junior accountant role. Criterion "month-end close": the candidate runs bank reconciliations for 40 entities in her current job. Gap: no payroll experience. She asked about hybrid work. Notice: four weeks. Salary: within the posted range. Recommendation: advance to the technical interview.

6. Account review summary template

Account review summary
Date: | Account: | Renewal date:
Attendees:
Results since last review (numbers):
What is working:
Problems raised:
Health: green / amber / red, with reason
Renewal or growth signals:
Commitments we made:
Commitments they made:
Next review:

Example (fictional): Quarterly review, Crestline Fitness. Results: 1,240 members booked classes in the app, up from 980. Working: automated reminders. Problem: the Android app is slow at peak times. Health: amber until the speed fix ships. Signal: they asked about two new locations. We fix performance by 30 June; they share opening dates.

Prompts that turn a transcript into each template

One master prompt works for all six templates. Paste the template, then a transcript with speaker names, so the model can assign owners.

You are writing a call summary for [team] at [company].
Fill in the template below using only facts stated in the transcript.
Rules:
- If a field was not discussed, write "Not discussed". Never guess.
- Copy names, numbers, prices and dates exactly as said.
- Every action item needs an owner and a due date. If none was
  said, write "Owner: unassigned" or "Due: not set".
- Keep each bullet under 25 words. No opinions about people.
- Put anything uncertain or contradictory under "Open questions".

TEMPLATE:
[paste template]

TRANSCRIPT:
[paste transcript with speaker names]

Then add one line for the call type:

• Discovery: "Quote the prospect's own words for each pain point."
• Support: "List troubleshooting steps in order."
• Kickoff: "Separate in-scope items from items ruled out."
• Internal: "List a decision only if someone clearly agreed."
• Recruiting: "Record evidence per criterion, not personality."
• Account review: "List every promise by either side, with its date."

For more prompts by job, see our 50 AI prompts for small business.

How to automate call summaries: record, transcribe, summarize, file

Every setup has four steps: record the call, transcribe it with speaker names, summarize the transcript into your template, and file the result. Use what your meeting platform includes, or build your own pipeline.

Built-in options in Zoom, Google Meet and Microsoft Teams

Platform	Feature	What you need (as of 18 October 2026)	Good to know
Zoom	Meeting summary with AI	Zoom Workplace Pro, Pro Plus, Business, Business Plus, Enterprise or Enterprise Plus	Admins can enable summary templates; not in breakout rooms
Google Meet	Take notes for me (Gemini)	Organizer on an eligible edition (such as Business Standard or Plus, Enterprise Standard or Plus, Frontline Plus) or Google AI plan	Notes go to a Google Doc in the organizer's Drive; eight languages
Microsoft Teams	Intelligent recap	Microsoft 365 with Teams, plus a Teams Premium or Microsoft Copilot license; transcription on	AI notes and suggested tasks; full recap needs recording allowed

Built-in summaries are the fastest start. Their limits: the platform's format, no knowledge of your CRM fields, and no phone calls.

A custom pipeline that files summaries in your CRM

1. Trigger: a "recording ready" event, or an upload for phone calls.
2. Transcript: Zoom saves cloud recording transcripts as VTT files (English only), the Google Meet REST API returns transcripts, and Microsoft Graph returns Teams meeting transcripts. Phone calls need a speech-to-text service.
3. Summarize: send transcript and template to a language model and ask for JSON that maps to your CRM fields.
4. Review: the call owner fixes and approves the draft.
5. File: post the JSON to your CRM, directly or through a webhook in n8n, Zapier or Make.

The Python example below covers steps 3 to 5 for discovery calls. The Anthropic SDK's messages.parse() returns output that matches a Pydantic model and raises a ValidationError when it cannot, for example after a refusal. The webhook URL and deal ID are placeholders.

import anthropic
import requests
from pydantic import BaseModel, ValidationError

WEBHOOK_URL = "https://example.com/your-crm-webhook"


class ActionItem(BaseModel):
    task: str
    owner: str | None
    due_date: str | None


class DiscoverySummary(BaseModel):
    attendees: list[str]
    pain_points: list[str]
    current_solution: str | None
    budget: str | None
    decision_process: str | None
    timeline: str | None
    objections: list[str]
    next_step: str | None
    action_items: list[ActionItem]
    open_questions: list[str]


RULES = (
    "Summarize this sales discovery call for our CRM. "
    "Use only facts stated in the transcript. Use null or an empty list "
    "for anything not discussed. Copy names, numbers and dates exactly."
)

client = anthropic.Anthropic()


def summarize(transcript: str) -> DiscoverySummary | None:
    try:
        response = client.messages.parse(
            model="claude-opus-5-5",
            max_tokens=16000,
            messages=[{
                "role": "user",
                "content": [
                    {"type": "text", "text": RULES},
                    {"type": "text", "text": transcript},
                ],
            }],
            output_format=DiscoverySummary,
        )
    except ValidationError:
        return None  # refused or cut off: write these notes by hand
    return response.parsed_output


with open("transcript.txt", encoding="utf-8") as f:
    draft = summarize(f.read())

if draft is not None:
    print(draft.model_dump_json(indent=2))
    if input("File this summary in the CRM? (y/n) ").strip().lower() == "y":
        payload = {"deal_id": "DEAL-1042", "summary": draft.model_dump()}
        requests.post(WEBHOOK_URL, json=payload, timeout=30).raise_for_status()

Model cost per call is small but adds up across a team; our LLM API cost guide shows how to estimate it. Then see CRM automation for small business for what to do with the data.

Recording consent: the short version

This is general information, not legal advice. The rules depend on where each participant is.

• United States: federal law (18 U.S.C. 2511) allows recording when one party consents. Some states require everyone's consent; the Digital Media Law Project, in a guide it no longer updates, lists California, Florida and Washington among them.
• United Kingdom: the ICO says people must be told how their personal data will be used when you collect it, including the purpose, lawful basis and retention period.
• Everywhere: announce the recording and the AI summary at the start, and let people object.

Google Meet tells all participants when notes are being taken, and Workspace admins can require explicit consent first. Zoom admins can show a custom message when AI is turned on. Put your rules in an AI acceptable use policy.

Accuracy checks before a summary is filed

AI summaries can be wrong in quiet ways; Google's help page warns that Meet notes can be incomplete or inaccurate. Check five things:

1. Numbers and dates: compare every price, quantity and deadline with the transcript.
2. Names and owners: transcripts often mishear names.
3. Commitments: each "we will" was agreed, not just suggested.
4. Missing fields: "Not discussed" shows what to ask next time.
5. Tone: no opinions about people.

Then track the share of summaries approved without edits. If it drops, check the template, prompt and transcript before you blame the model.

Want summaries in your own templates, checked and filed in your CRM? See our AI meeting notes and call summaries tool, or tell us about your calls. We quote a fixed price, and you own the code.

AI call summary FAQ

What is an AI call summary?

An AI call summary is a short written record of a call, produced by a language model from the transcript. It lists the purpose, key points, decisions, action items with owners and dates, and open questions. Good summaries follow a fixed template for each call type and include only facts that were said, so someone who missed the call can act on it.

How do I summarize a meeting with AI?

Record the meeting with everyone's knowledge and get a transcript with speaker names. Give a language model the transcript, a template and strict rules: use only stated facts, copy numbers exactly, and mark missing items as not discussed. Check names and dates before you share it. Zoom, Google Meet and Microsoft Teams can write the first draft on eligible plans.

What should AI discovery call notes include?

Discovery call notes should capture the prospect's pain points in their own words, their current solution, the budget if it came up, who decides and how, the timeline and its trigger, objections, and a dated next step. These fields map neatly to CRM properties, so managers can compare deals without listening to recordings.

Are AI call summaries accurate enough for sales teams?

They make a strong first draft, not a final record. Errors usually come from the transcript, such as misheard names, numbers or crosstalk, or from a model filling a gap with a guess. Strict prompts, structured output and a two-minute review by the rep keep those errors out of the CRM. Track how often reps approve summaries without edits.

Should I use built-in meeting summaries or a custom tool?

Built-in summaries in Zoom, Google Meet or Teams are the quickest start if your plan includes them and a general format is enough. A custom pipeline makes sense when you need your own template per call type, phone calls covered, summaries written into specific CRM fields, or recordings kept under your control.

Sources

• Zoom Support: Meeting summary with AI
• Zoom Support: Audio transcription for cloud recordings
• Google Meet Help: Take notes for me
• Google Workspace Updates: Take notes for me settings (July 2026)
• Google Workspace Updates: Explicit consent in Google Meet
• Google for Developers: Meet REST API artifacts
• Microsoft Learn: Intelligent recap in Teams
• Microsoft Learn: Graph callTranscript resource
• Anthropic: Structured outputs
• Cornell LII: 18 U.S. Code 2511
• Digital Media Law Project: Recording calls and conversations
• ICO: Right to be informed