1,600 calls a month, without retyping a word

Project Overview

Who is Immo Cauwe?

Immo Cauwe is a real estate agency from Knokke with a large property management division. The agency manages an extensive portfolio of apartment buildings along the coast.

The phone is its most important channel. Owners, tenants and suppliers call all day about a leak, a key, an invoice or a general assembly. In June 2026, that added up to around 1,600 calls in a single month.

Every call then had to be processed by hand. A property manager typed up a note, looked up the right building and file in Keypoint, and created the tasks. That takes a lot of time, and information gets lost along the way.

The challenge

"How do we make sure every phone call automatically lands in the right file, without a property manager having to retype everything?"

There was a second question too. A large share of calls aren't urgent and could easily wait until the next working day. Yet today every call is still handled by a person — by an employee during office hours, and by the call center outside of them.

Project Delivery

The approach

We started with a process analysis of the base case: an incoming call from A to Z. Together with Immo Cauwe, we mapped out how the phone system (3CX), the transcription (Leexi) and the property management software (Keypoint) relate to each other. That led to a phased plan. First get call processing reliable. Then clean up the data. Only then automate the phone itself.

We work deliberately with small, working versions — go live fast and adjust based on real calls, instead of spending months building something that hasn't delivered anything yet.

What we built

1 · ImmoLink: from call to file in Keypoint

ImmoLink is the integration that automatically turns every phone call into a report and a file.

1. The call comes in through 3CX and gets transcribed by Leexi.
2. Based on the phone number, the system looks up the caller in Keypoint: the owner or company, the building and the unit.
3. An A.I. agent writes a short report in bullet points, with the key details and suggested tasks.
4. The property manager gets an email with a link to a screen where they review the report, adjust tasks, and pick the file or create a new one.
5. After confirmation, everything goes automatically to Keypoint and SharePoint: the report as a document, and the notes and tasks attached to the right file.

The benefit: a property manager no longer has to retype anything, and every call can be found back in the file.

Human in the loop. Not every number is instantly recognizable. If a number belongs to multiple owners, or to none at all, the property manager gets a selection form instead of a wrong match. That keeps the data in Keypoint clean. We used n8n for the first version and then built a dedicated web app, because that's faster to adjust and doesn't need to run everything through email.

The rollout happened in small steps: first a test group of three property managers, then wider.

2 · Pulling phone numbers out of notes

During the first months, we saw that 35 to 40 percent of calls couldn't be matched automatically. The reason: a lot of phone numbers in Keypoint aren't in the phone field, but buried in the free-text notes field.

So we built an A.I. script that reads through all the notes, recognizes phone numbers, normalizes them, and prepares them to be attached to the contact as an actual phone number. Context like "rented to" gets picked up too and added as a label. The notes themselves stay unchanged, and existing labels don't get overwritten.

We started with a sample of 100 contacts out of a file of about 8,000, so Immo Cauwe and Keypoint could validate the quality first. Only after that does the full run happen.

The benefit: the more numbers are correctly matched, the more calls land automatically in the right file.

3 · Triage bot: answering the phone itself

The next step is an A.I. voice agent that handles first-line calls. The bot recognizes the caller, lets them explain what's going on, summarizes it, and decides what happens next:

• Clearly not urgent, like an invoice or a certificate: the bot confirms it'll be handled the next working day, sends an email, and parks the file in Keypoint through the existing ImmoLink flow.
• Clearly urgent, like a gas smell or fire: escalate immediately to a property manager or the call center.
• Unclear cases: the bot asks four questions covering danger, damage, safety and nuisance, and assigns a priority from P0 to P4.

Key principle: the caller always decides. The moment someone objects, the bot transfers the call right away. Since every call still reaches a person today anyway, the bot can only reduce the volume, never make things worse. Every disagreement between the bot and the caller gets logged and used to fine-tune the triage.

Ongoing collaboration

Alongside the build, there's a maintenance and support setup running with fixed response and resolution times, so the integration keeps working and small changes can happen quickly.

Results

Phone calls today get automatically transcribed, summarized and prepared as a file in Keypoint. The property manager checks and confirms, instead of typing and looking everything up themselves. Reports, notes and tasks land in the right place immediately, so colleagues can pick up a file without first having to call around.

By cleaning up around 8,000 contacts, we're tackling the biggest bottleneck: the 35 to 40 percent of calls that still don't get matched automatically today. Immo Cauwe itself estimates that around 40 percent of calls are routine matters that a bot could handle. That's the foundation for the triage bot now ready to go.

Flowkify keeps supporting Immo Cauwe as its A.I. partner. The next steps are finishing the contact data clean-up and rolling out the voice agent, starting during office hours on closed lines.

Curious what A.I. could mean for your real estate or property management firm? Get in touch with Flowkify for a tailored A.I. scan.