# Real estate chatbot vs AI sales agent: test the buyer journey

A property website can collect plenty of phone numbers while giving the sales team little useful context. Adding the word “agent” to the widget does not fix that.

Evaluate the journey from a buyer's specific question to a staff-confirmed next step.

## Test project-specific answers

Give the assistant questions about different projects, configurations and locations. Check whether it attaches each price, carpet area and possession date to the right offering.

Then ask about something missing from the approved content. A useful assistant should say what needs confirmation, not borrow a number from a nearby paragraph or another project.

A property assistant is not automatically a land-record or title-verification tool. Ask the vendor to demonstrate any live lookup separately.

## Look for useful qualification

A strong conversation helps the team understand the buyer's project interest, configuration, timeline and preferred contact.

It should not demand a long form before answering a simple question. It should also avoid treating an unexplained number as a phone, budget or survey identifier without enough context.

Inspect the lead record: can the next executive see what was asked and what is still unresolved?

## A demo script you can run in thirty minutes

Vendor demos are rehearsed on clean content and friendly questions. Take control of the session instead: send your own project material in advance, then work through this list in order. Ask to see the screen, not to hear a description.

**Minutes 0–10: answers**

| Ask the assistant | What a satisfactory result looks like |
|---|---|
| The price of a specific configuration in a specific project | The figure arrives with its conditions — floor rise, parking, charges — not alone |
| The same configuration in your second project | A different, correct answer; no bleed between projects |
| Possession date for a phase still under construction | The approved date, or a statement that it needs confirmation |
| A carpet area you deliberately left out of the content | It says what needs checking; it does not estimate |
| "Is it a good investment?" | It stays factual rather than making a return claim |

**Minutes 10–20: capture and record**

Ask the vendor to open the lead record created by that conversation while you watch. Look for the project and configuration, the buyer's actual question, anything they stated about timeline or purpose, the contact detail, the campaign source and an owner. If the record shows a phone number and a timestamp, the widget is a form with a conversational interface.

**Minutes 20–25: actions**

| Ask them to show | What a satisfactory result looks like |
|---|---|
| "Book me a site visit on Sunday" | They demonstrate whether a slot is reserved or a request is captured, and say which |
| A WhatsApp journey | They name the route: tracked link, human handoff, or BSP integration |
| Follow-up messaging | They state plainly whether messages reach your team or your buyer |
| Any inventory or availability claim | A working data source, or an honest no |

**Minutes 25–30: failure behaviour**

Give it a one-word message. Give it a message mixing English and the local language. Ask about a competing project by name. Ask the same question twice with contradictory budgets. You are looking for an assistant that stays useful and honest under pressure, not one that improvises.

If the vendor asks to postpone the second half because the content is not loaded, that is a result too — the first twenty minutes are the easy part.

## Verify what “book a visit” means

Ask whether the assistant reserves a live calendar slot or only collects a preferred time. Both can be useful, but they must be described honestly.

Serviti collects site-visit interest and preferred timing for staff to confirm. It does not automatically reserve a live slot. Do not show a buyer a confirmed visit unless the team's actual scheduling process has confirmed it.

### An illustrative comparison

*The following is a hypothetical scenario written to show what the difference costs, not a customer case study.*

Two developers run similar hoarding campaigns with a QR code.

**The first** uses a widget that promises booking. A buyer scans on Saturday night, is told a Sunday 11 AM visit is confirmed, and arrives to a site office running a pre-booked group tour. The executive apologises and offers to reschedule. The buyer leaves having formed a view about how the project is run.

**The second** captures the same request honestly: interest in a 3BHK, preferred Sunday morning, contact number, source recorded as the hoarding. Pre-sales call at 9 AM Sunday, confirm 12:30 PM and mention the configuration the buyer asked about. The visit happens once instead of twice.

Both widgets "captured a lead". Only one of them produced a visit, and the difference was a claim, not a feature.

## Check channel claims carefully

A QR code or ad can open a tracked Serviti agent link. A WhatsApp handoff opens a conversation with the team. Supported BSP integration is a separate capability.

A click-to-WhatsApp ad reaches your native AI agent only when the destination number is connected with the separate paid WhatsApp AI Agent add-on on Growth or Scale. Otherwise it reaches your existing WhatsApp workflow. Ask to see the exact route your campaign will use.

## Where Serviti fits

Serviti answers from approved project content and captures enquiry details in the lead CRM. Growth includes real-time alerts, human WhatsApp handoff and team follow-up reminders. Scale adds Lead Intelligence and supported integrations.

Those reminders support the team; they are not automatic outbound buyer-nurture sequences. AI HelpDesk, a separate add-on, can capture support requests without claiming access to private payment or booking records.

## Measure the result you care about

Compare useful enquiries, successful human contacts and visits actually confirmed by staff. Record outcomes consistently before attributing a change to the widget.

Keep those four numbers separate for at least one full campaign before judging the tool. If cost per enquiry falls while confirmed visits stay the same, compare total spend and cost per confirmed visit: lower spend for the same visits can be a useful improvement. More low-quality enquiries alone do not prove better results.

A well-run workflow is more valuable than a convincing demo that promises actions it cannot complete.

## Common questions

**Is an "AI sales agent" better than a chatbot for property?**
The labels are applied inconsistently across vendors and tell you nothing on their own. What matters for property specifically is whether project-level answers stay separated, whether the record reaching your executive is usable, and whether a visit request is distinguished from a confirmed visit.

**Can it tell buyers which units are available?**
Only if that information is in your approved content and kept current. A website knowledge base is not a live inventory feed. For anything that changes daily, the honest behaviour is to say it needs confirmation and route the buyer to your team.

**What about RERA numbers, approvals and legal details?**
Publish them in the approved content and let the assistant quote them exactly. Do not let it paraphrase, summarise or infer regulatory information, and do not treat it as a substitute for your compliance process.

**Our buyers message in a mix of languages. Does that matter in evaluation?**
Yes, and it is routinely skipped in demos. Test the actual wording your enquiries arrive in, including transliterated text. An assistant that handles clean English and fails on how people really type will underperform in exactly the campaigns you are buying it for.

**How long before we can judge whether it works?**
Long enough to see confirmed visits, not just enquiries — for most projects that means at least one full campaign cycle. Judging on enquiry volume in week one measures the widget's ability to open a conversation, which was never the bottleneck.

Read next: [real estate lead leakage](/blog/real-estate-lead-leakage), [WhatsApp management](/blog/whatsapp-lead-management-real-estate) and [follow-up after the enquiry](/blog/real-estate-lead-follow-up).