The Voice Note That Updates Your CRM for You

Wednesday, 4.47pm. A viewing has just finished. The applicant likes the house, but not the kitchen. They are chain-free, have a mortgage agreement in principle and want to discuss an offer with their partner that evening.

The negotiator makes a quick voice note in the car park: “Sarah and Tom liked the house overall, concerned about the kitchen, probably around asking price, need a call tomorrow morning.”

Then the next viewing starts.

By 5.30pm, the note is still on the phone. By 6pm, it is buried under WhatsApp messages, missed calls and another viewing. Nothing has been added to Reapit, Alto or Dezrez. The vendor sees no update. The applicant receives no follow-up. The useful detail is still in somebody’s head.

I spent ten years on the agency floor. I know how this happens. It is rarely because a negotiator does not care. It happens because the gap between a real conversation and a clean CRM record is wider than most principals realise.

Speed-to-lead does not stop after the first enquiry

Most agencies talk about speed-to-lead as the time between a Rightmove enquiry and the first call. That matters, but it is only the first part of the job.

The same principle applies after a viewing.

The sooner you capture feedback, the more accurate it is. The sooner the applicant receives a relevant response, the more likely they are to stay engaged. The sooner the vendor sees a proper update, the more confidence they have in the agency handling their instruction.

The problem is that a viewing note is often unstructured. It might say:

“Good viewing. Likes it. Call tomorrow.”

That is not enough for the next person picking up the applicant record. It does not explain what they liked, what held them back, whether the applicant is proceedable, or what action is expected next.

A good negotiator may remember all of that. A busy branch manager looking at the CRM two days later will not.

This is where a spoken voice note can become useful. Not as a replacement for judgement, and not as another separate system, but as a quicker way to get the right information into the record while the conversation is still fresh.

What should actually be captured?

A voice-to-CRM setup should not simply turn every spoken word into a long block of text. That creates a different admin problem.

The useful information needs to be separated into the fields and records your team already uses. For a viewing, that might include:

• The applicant or buyer
• The property and viewing date
• Whether the viewing went ahead or was a no-show
• Overall level of interest
• Positive feedback
• Objections or concerns
• Whether the applicant is proceedable
• Whether an offer was mentioned
• The next action
• Who owns that action
• When it needs to happen

A negotiator might say:

“Viewing at 18 Oak Road with Sarah Malik and Tom Evans. They liked the location and the garden. The kitchen is putting them off, so they want to talk it through tonight. They are chain-free and have an agreement in principle. Call Sarah at 9.30 tomorrow. Do not describe this as an offer.”

The system should turn that into a clear activity note, not invent an offer.

Depending on the CRM and the way it is configured, the information could land in the applicant’s timeline, the property activity record, a viewing feedback section or a follow-up task. The exact destination matters.

Reapit, Alto and Dezrez do not use identical field structures. Some agencies have heavily customised stages, tags and activity types. Others rely mainly on free-text notes. Before anything is connected, we need to understand where your team actually looks for information.

If a negotiator records the note against the wrong property, the process has failed even if the transcription is perfect.

The important bit is field mapping

This is the part that is often skipped.

A voice note is unstructured. Your CRM is not. It expects a person, a property, an event, a status and, often, a task. Those details need to be mapped properly.

For example:

• “Sarah Malik” needs to match the correct applicant record.
• “18 Oak Road” needs to match the right instruction.
• “Likes it but concerned about the kitchen” may belong in feedback, not the offer field.
• “Call tomorrow at 9.30” needs to create a task with an owner and due time.
• “Chain-free” may update a qualification field, but only if your team wants that information recorded there.

Names are a particular problem. Estate agency teams work with similar names, multiple applicants and couples sharing a phone number. Addresses can also be misheard, especially in busy cars or on noisy high streets.

That is why a sensible setup uses identifiers wherever possible: the viewing reference, the mobile number, the applicant record already open on the negotiator’s phone, or a property reference spoken at the beginning of the note.

The instruction to the team might be simple:

“Start every note with the property reference and applicant name.”

That small habit can make the difference between a reliable record and a dangerous guess.

What still needs a human check?

I would not allow a system to write everything directly into a live CRM without sensible checks.

A human should still review:

• Unclear names or addresses
• Whether the person is the applicant, vendor, tenant or solicitor
• Any mention of an offer
• Price-sensitive conversations
• Mortgage or chain details
• Complaints or sensitive personal information
• The assigned negotiator
• The wording of the vendor update

The system can prepare the record. It should not decide what the applicant “really meant”.

If someone says, “They could possibly stretch a bit, but they need to speak to the broker,” that is not an offer. It is not permission to change a price field. It is context for a negotiator to follow up.

The right approach is usually one of three options:

1. Write low-risk notes automatically.
2. Send higher-risk updates to a review queue.
3. Ask the negotiator to confirm before anything changes a formal status.

That keeps the speed while leaving judgement where it belongs.

GDPR and the practical details

A spoken note about an applicant is personal data. If it contains mortgage information, family circumstances or other sensitive details, the controls need to be clear.

The agency should know:

• Where the audio is stored
• Whether the original recording is retained after transcription
• Who can access the transcript
• How long records are kept
• Which suppliers process the information
• How the note is corrected if it is wrong

The team should also know when a live customer call is being recorded or processed. An AI voice agent answering inbound calls needs a clear disclosure at the start of the call, and your privacy information should explain how recordings and transcripts are used.

For staff voice notes, the main principle is data minimisation. Capture what is needed to progress the applicant, update the vendor and create the next action. Do not create a permanent audio archive simply because storage is available.

How we approach it at Nexform AI

At Nexform, we start with the agency’s real process, not a generic demonstration.

I want to see how a viewing is booked, where the feedback is currently recorded, how Reapit, Alto or Dezrez is configured, and what the branch manager expects to see before the end of the day.

We then map the fields, test sample notes and check what happens when the audio is unclear or the applicant cannot be matched confidently.

The usual route is a 30-minute Discovery call, followed by a paid Pilot priced between £3,000 and £8,000. The Pilot tests one specific process with real examples. If it proves useful, we move to a wider Build. If it does not, we stop with a clear answer rather than adding another tool your team has to work around.

If viewing feedback is sitting in phones, notebooks and half-finished CRM notes, book a 30-minute Discovery call. We will look at where the information currently gets lost, what should be written into Reapit, Alto or Dezrez, and what a sensible Pilot would cover.

No grand presentation. Just a practical conversation about how your branch actually works.

About Phil Yassein

Phil Yassein is the founder of Nexform AI. He spent 10 years in UK estate agency across valuations, sales progression and branch operations before building practical AI automation for the problems he experienced first-hand. Nexform AI is an approved Crown Commercial Service supplier.