When an AI assistant follows up your property leads, the minimum useful record is one event row for every state change in an inquiry's life: arrival, automated response, delivery status, qualification decision, escalation, agent takeover and outcome. Each row carries a timestamp, what happened, who acted where someone did (the assistant or a named person), and a reason code where a decision was made. Those seven events let you rebuild an inquiry weeks later and trace where a follow-up failed, without storing the whole conversation.
This is the event-level record for one workflow: property lead follow-up. Several rows exist per inquiry, and they roll up into the generic five-field run log in our guide to monitoring an AI workflow after launch. Review cadence, escalation-rate thresholds and cost tracking live on that page, not here.
The seven events to record
An inquiry moves through eight stages from arrival to final outcome, mapped in our inquiry-to-agent workflow guide. Seven of those state changes are worth a record of their own:
| Event | Minimum fields |
|---|---|
| Inquiry arrival | Timestamp, channel, listing reference, opt-in status where the channel requires one. No actor: the lead starts this one |
| Automated response | Timestamp, actor recorded as automated, reference to the message sent |
| Delivery status | Timestamp, status the channel reported, error code if it failed |
| Qualification decision | Timestamp, actor, decision, reason code |
| Escalation | Timestamp, actor, reason code, plus the four escalation fields your fallback rules already define: unedited message, channel, listing reference, unverified gap |
| Agent takeover | Timestamp, actor recorded as human, agent's name |
| Outcome | Timestamp, result, reason code where one applies, actor who closed the inquiry |
Each follow-up attempt is its own automated response row, so a lead nudged three times across a week has three of them, and 'no response' is a legitimate outcome recorded on the same terms as a booked viewing. Record references to messages, not full transcripts by reflex. The same rule governs escalations in our fallback rules for property inquiries the AI cannot answer.
How to tell automated actions from human ones
Every event where someone acts names the actor: automated, or human with a name. Two rows carry no actor: arrival, started by the lead, and delivery status, reported by the channel. Qualification and escalation take an actor too: an agent can escalate a lead the assistant was still working. During a pilot this shows which decisions the assistant made on its own; in a dispute, whether a promise came from the assistant or a named agent. Make it a structured field, never a note in free text. Which actions the assistant may take at all is policy, covered in our approval rules for AI lead follow-up.
Reason codes, not free text
Free-text reasons hide patterns. A run of escalations noted as 'various issues' tells you nothing. Twenty rows coded missing property information tell you to fix your listings. Start with five codes:
- Escalated. Reuse the trigger categories from your fallback rules.
- Qualified. The lead met your criteria and is ready for an agent. The same code covers the qualification decision and the escalation it triggers.
- Delivery failed. WhatsApp's Business Platform reports outbound message status as sent, delivered, read or failed, with error details on failure, per Meta's message status reference. Whether they reach your record depends on the channel setup, so confirm with your provider.
- Missing property information. The assistant could not answer because the listing lacks the detail.
- Opt-out. The WhatsApp Business Messaging Policy only permits messaging people who provided their number and opted in. Record opt-in on the arrival row, and record an opt-out as an outcome row carrying this code, since it ends the automated follow-up. Never leave a stop request buried in message history.
A modelled event sequence for one inquiry
This sequence is modelled for illustration and is not drawn from any client system.
- 09:12: inquiry arrives on WhatsApp about a listed cottage. Channel, listing reference and opt-in recorded.
- 09:12: automated response sent. Actor: automated.
- 09:13: channel reports the message delivered.
- 09:38: qualification decision. Budget and moving date confirmed. Actor: automated. Reason code: qualified.
- 09:39: escalation to the duty agent. Actor: automated. Reason code: qualified.
- 10:05: agent takeover. Actor: human, agent named.
- 12:20: outcome recorded. Viewing booked. Actor: human.
Who owns the log
Give log completeness and accessibility to one named role: the sales manager, or the operations lead in a larger agency. This mirrors the role split in the approval rules, where the principal sets policy and the sales manager operates it. That person checks the log is actually being written, chases the gaps, and can pull an inquiry's full history when someone asks for it.
Data-controller accountability is a heavier question that sits above the log owner. Simon's guide to identifying your data controller before AI touches personal data covers who answers for the processing itself.
Retention and access
Inquiry messages contain personal data, and processing it in Zimbabwe falls under the Cyber and Data Protection Act, enforced by POTRAZ. Commentary from DLA Piper Africa and Manokore Attorneys notes that data controllers are responsible for the acts of their agents and representatives. Ask your lawyer how that responsibility applies when an assistant sends messages on your behalf. We do not publish a recommended retention period for lead-communication records. Set one with your lawyer, and decide who may access the log. Veritas reports that POTRAZ compliance inspections started on 1 September 2026 with nine priority sectors that do not include real estate. Ask your lawyer where your agency stands rather than assuming that holds.
Confirm the fields with your provider
Which of these fields your platform and channels actually surface is an onboarding question. Take the seven events into a lead-nurturing walkthrough and ask how each one appears in your records. See current pricing for what the Real Estate Platform includes. If you plan a shadow-mode run before going live, this record is what you will review, so settle it before the pilot starts.
See how the platform records your leads
TechTribe's Real Estate Platform includes AI lead nurturing, automated follow-up and a central lead dashboard. See how it records each stage of an inquiry.
About the author
Simon
Simon writes about websites, lead capture, and digital growth for real estate agencies in Zimbabwe.
