AI call notes
Also called AI meeting notes, AI notetaker, conversation intelligence, call transcription to CRM, AI call summaries
AI notetakers transcribe sales calls and push summaries into your CRM. Studies show real time saved on notes; accuracy and consent need care.
AI notes save real documentation time, though the best independent evidence comes from medicine, not sales. In a study of 46 clinicians, time in notes per appointment fell from 10.3 to 8.2 minutes and after-hours work fell by about 15 minutes a day. UK civil servants using an office AI assistant self-reported saving 26 minutes a day. Errors are rare but real: about 1% of transcriptions in one test contained invented sentences. Recording consent rules differ between South Africa and California.
At a glance
| Status | In pilots |
|---|---|
| What it does | Joins or records sales calls and video meetings, transcribes them, writes a summary with next steps, and saves it against the contact or deal in the CRM. |
| Typical cost | $19.99 per user per month for one vendor's business plan, which syncs notes to Salesforce or HubSpot (USD, billed annually, vendor list price, checked Oct 2026) |
| Data it needs | Calendar access, call or video-meeting audio, and a CRM where contacts and deals already exist so each note can be matched to the right record. |
| Setup effort | Low: connect calendar and CRM, add recording notices, and decide which meetings are recorded. |
| Evidence | Moderate. One independent clinical study of an AI scribe, a large UK government trial and a maker-run field experiment; no independent sales-specific study. |
What are AI call notes in the CRM?
After a sales call someone has to write down what the client needs, what was agreed and what happens next, then type it into the CRM. Many salespeople skip it, so the CRM goes stale and the owner cannot see the pipeline. A vendor survey of 4,050 sales professionals found the average seller spends only 40% of their time actually selling.
AI notetakers go after that admin. A bot joins the video call, or the phone system records the call; software transcribes the audio, a language model writes a summary and action list, and an integration attaches it to the right contact and deal. Some tools also update CRM fields such as deal stage or next-step date. Because transcription and summarising are well-understood tasks, this is one of the more mature uses of AI in sales.
How does it work?
- Capture. The tool reads your calendar and sends a bot into Zoom, Google Meet or Teams calls, or records phone calls through your calling system. In-person meetings need a phone app.
- Transcribe. Speech-to-text turns audio into text and labels who spoke.
- Summarise. A language model writes a summary, decisions, objections and next steps, often from a template you set.
- Push. The summary is written to the matching CRM record. One vendor's business plan lists synced notes to Salesforce or HubSpot.
- Review. The salesperson checks and corrects the note before it becomes the record.
The last step matters more than it looks. In the clinical study below, some clinicians found the AI notes accurate and detailed, while others said checking and fixing them took as long as writing their own.
What it needs to work
Call notes only help if each call can be matched to a clean CRM record, and everyone on the call knows it is being recorded.
- Calendar invites that include client email addresses, so the tool can match each meeting to an existing CRM contact and deal
- A CRM with de-duplicated contacts and one open deal per opportunity, so notes attach to the right record
- Calls held on a supported platform (Zoom, Google Meet, Teams or a recording phone system), not personal mobiles
- A recording notice in meeting invites and in your privacy notice, and a simple way for a client to ask you not to record
- A short note template the sales team agrees: need, budget, decision-maker, objections, next step and date
- A rule that the salesperson reviews and corrects each AI note within one working day
Most use cases in this guide succeed or fail here, on the data underneath them, before any AI is involved.
What does the evidence show?
We found no independent study of AI call notes in sales teams, so the evidence comes from neighbouring fields.
In healthcare, where note-taking time is closely measured, a 2025 study followed 46 outpatient clinicians using an AI scribe that listens in the background. Time in notes per appointment fell from 10.3 to 8.2 minutes, and after-hours work fell from 50.6 to 35.4 minutes per scheduled day. It was a before-and-after study without a control group.
A UK government experiment gave 20,000 civil servants an office AI assistant for three months. Respondents reported saving 26 minutes a day on average, and Teams, where meetings happen, was the most-used app. Those savings are self-reported. A field experiment across 66 large firms, run by the tool's maker, found the same kind of assistant cut email time but did not change time spent in meetings. The time saving on notes rests mainly on healthcare evidence; whether better notes win more deals has not been measured.
AI call notes vs salesperson-written notes
| Compared on | AI call notes | Salesperson-written notes |
|---|---|---|
| Time per call | A few minutes to review and correct | Longer to write, and often skipped or done days later |
| Completeness | Captures everything said, including details the rep missed | Captures what the rep judged important |
| Error profile | Occasional mishearings and invented sentences in a confident tone | Omissions and optimism about the deal |
| Cost model | Per user per month subscription | Salesperson time |
| Best for | Teams with many calls and patchy CRM discipline | Low call volumes or highly confidential conversations |
Status and evidence
Rated pilots: off-the-shelf notetakers with CRM sync are sold at public per-user prices, but none of the measured evidence comes from sales work. One independent clinical study found less time spent on notes, a UK government trial found self-reported time savings, and the large field experiment was run by the tool's maker. Evidence is moderate rather than strong because none of these studies is about sales calls or CRM records.
Where does it go wrong?
- Recording without proper notice. South Africa's RICA lets a party to a call record it, but POPIA still requires you to tell people what you collect and why. California makes recording a confidential conversation without every party's consent a crime, with fines of up to $2,500 per violation. A 2025 US class action accuses one notetaker of recording meetings without all participants' permission and using recordings to train its AI.
- Invented content. In one test of a widely used speech-to-text model, about 1% of transcriptions contained whole phrases or sentences never spoken, and 38% of those included harmful content.
- Notes on the wrong record. Duplicate contacts in the CRM send summaries to the wrong deal.
- Data leaving the country. Cloud tools may process audio abroad, which POPIA section 72 allows only on listed grounds.
- Nobody reads them. Long summaries pile up; the CRM fills with text while next steps still slip.
Numbers it should move
Ask the demo business
Northgate Supply is a fictional wholesale distributor of catering and packaging supplies: 28 staff, R46.6m in sales over the last year. Its data is generated for this demo, to 30 September 2026.
Half of September's 180 customer queries got a first reply within 2.2 hours, but 3 waited more than a working day. December was the slowest month, at 4.7 hours.
How this was worked out
Median hours from a customer query arriving to the first reply, per month, across email and WhatsApp.
Where the data usually lives
Where the records this use case needs are usually kept in a small business. It is not a list of recommended tools.
Frequently asked questions
How much does an AI notetaker that syncs to the CRM cost?
Public prices start low. One vendor's business plan is $19.99 per user per month, billed annually, and includes syncing notes to Salesforce or HubSpot (USD, vendor list price, checked Oct 2026). For a five-person sales team that is about $100 a month. Some CRMs also bundle transcription into higher tiers. The larger cost is the time salespeople spend checking each note.
Do I need IT support to set up AI meeting notes?
Usually not. Most tools connect when you sign in with your Google or Microsoft account and authorise your CRM. The work is policy rather than technology: deciding which calls are recorded, adding a recording notice, setting how long recordings are kept, and checking that notes land on the right CRM records. A Google Workspace or Microsoft 365 administrator may need to approve calendar access.
Is it legal to record sales calls with AI in South Africa and the UK?
In South Africa, RICA allows a party to a conversation to record it, but POPIA requires you to tell people what you collect, why and who receives it, and restricts sending personal information abroad. In the UK, UK GDPR requires clear privacy information when you collect personal data. In California, all parties must consent to recording a confidential conversation. See the Information Regulator and the ICO. This is not legal advice.
How accurate are AI call summaries?
Accurate enough to save time, not accurate enough to skip review. Clinicians in one study reported mixed accuracy, from notes they trusted to notes that took as long to fix as to write. A test of one widely used speech-to-text model found about 1% of transcriptions contained phrases nobody said. Check names, numbers, prices and commitments before a note becomes the record.
Will AI notes make my salespeople sell more?
They cut admin, which is the clearest gain. A vendor survey reports the average seller spends only 40% of their time selling. In a maker-run experiment, an office AI assistant cut email time but did not change time in meetings. Whether reps turn saved minutes into more calls is a management question; no independent study has measured it in sales.
More in sales & new business
Sources
- 1Clinician Experiences With Ambient Scribe Technology to Assist With Documentation Burden and EfficiencyJAMA Network Open, 2025Study
- 2Careless Whisper: Speech-to-Text Hallucination HarmsarXiv (Koenecke, Choi, Mei, Schellmann, Sloane), 2024Study
- 3Microsoft 365 Copilot Experiment: Cross-Government Findings ReportGovernment Digital Service (UK), 2025Government
- 4Shifting Work Patterns with Generative AIarXiv / NBER Working Paper 33795 (Dillon, Jaffe, Immorlica, Stanton), 2025Vendor
- 5
- 6Pricing | Otter.aiOtter.ai, 2026Vendor
- 7Regulation of Interception of Communications and Provision of Communication-Related Information Act 70 of 2002Department of Justice and Constitutional Development (South Africa), 2002Government
- 8California Penal Code section 632State of California (text via california.public.law), 2017Government
- 9
- 10Section 18 Notification to data subject when collecting personal information - POPIAProtection of Personal Information Act 4 of 2013 (text via popia.co.za), 2013Government
- 11Section 72 Transfers of personal information outside Republic - POPIAProtection of Personal Information Act 4 of 2013 (text via popia.co.za), 2013Government
- 12The right to be informedInformation Commissioner's Office (UK), 2024Government
This entry is educational. It is not legal, financial or professional advice, and it does not recommend any vendor. Figures attributed to vendors are their own claims.
- 1Clinician Experiences With Ambient Scribe Technology to Assist With Documentation Burden and EfficiencyJAMA Network Open, 2025Study
- 2Careless Whisper: Speech-to-Text Hallucination HarmsarXiv (Koenecke, Choi, Mei, Schellmann, Sloane), 2024Study
- 3Microsoft 365 Copilot Experiment: Cross-Government Findings ReportGovernment Digital Service (UK), 2025Government
- 4Shifting Work Patterns with Generative AIarXiv / NBER Working Paper 33795 (Dillon, Jaffe, Immorlica, Stanton), 2025Vendor
- 5
- 6Pricing | Otter.aiOtter.ai, 2026Vendor
- 7Regulation of Interception of Communications and Provision of Communication-Related Information Act 70 of 2002Department of Justice and Constitutional Development (South Africa), 2002Government
- 8California Penal Code section 632State of California (text via california.public.law), 2017Government
- 9
- 10Section 18 Notification to data subject when collecting personal information - POPIAProtection of Personal Information Act 4 of 2013 (text via popia.co.za), 2013Government
- 11Section 72 Transfers of personal information outside Republic - POPIAProtection of Personal Information Act 4 of 2013 (text via popia.co.za), 2013Government
- 12The right to be informedInformation Commissioner's Office (UK), 2024Government