Gains:
- Ability to configure the buyer/tenant need profile with artificial intelligence and intelligently match it with the portfolio
- Ability to avoid illegal filtering by observing discrimination, prejudice and KVKK limits in matching
- Ability to use the AI matching recommendation as a shortlist for the consultant to review, rather than as a final one
The heart of real estate is matching: matching the right property with the right buyer. An advisor may have dozens of properties in his portfolio and dozens of buyers on his list. Who deals with which property? Whose budget is enough for what? Who will actually buy it, who is just wandering around? Answering these questions quickly and accurately saves time and sales. Artificial intelligence is a powerful assistant here: it translates scattered customer notes into a structured needs profile, compares it with the portfolio, shortlists the 3-5 most suitable properties, even writes a draft of the recommendation message to be sent to the customer. But this unit has two critical caveats: matching cannot be based on discriminatory criteria, and the AI's recommendation is not the final decision but a shortlist for the consultant to review.
Needs analysis: what does he really want?
Customers often do not say exactly what they want; Those who say "a house with a garden" actually want "a place that is safe for children, has open space, and is close to the school." A good needs analysis reveals the real need underneath, not the demand on the surface. This requires structured criteria: budget range and financing (cash/credit), number of rooms and m², location preferences (proximity to work/school/transportation), moving time, must-haves (elevator, parking, furnished), what can be flexible, and “why buy now” motivation.
AI can put the (anonymous) notes of your conversation with the customer under these headings and remind you of the missing questions. This way, you truly understand the buyer without missing any critical information.
Tip: Separate “must have” from “nice to have” in the needs profile. Most buyers want it all, but the budget won't accommodate it. Give these two lists separately to the AI; The shortlist is thus realistic.
Critical limit: discrimination and KVKK
Matching must be based only on legitimate and property-related criteria: budget, number of rooms, location, m², financing availability, move-in time. Eliminating buyers/tenants based on criteria such as religion, language, ethnicity, marital status, political opinion, disability is both unlawful and discriminatory. AI can magnify bias in data: if there is a pattern in historical data that eliminates certain groups, the model can “learn” this and repeat it. Therefore, you should determine the matching criteria and not give any discriminatory fields to the model.
Additionally, customer data is within the scope of KVKK: the recipient's income information, family status, and contact information must be protected. When matching, give the AI only the necessary criteria, anonymously.
Attention: A statement such as "This type of family would suit this neighborhood" is discrimination and is prohibited, even if it seems well-intentioned. The buyer decides which option the buyer prefers; You show suitable properties according to legitimate criteria.
Step by step: pairing flow
- Configure the meeting note: Give the anonymous note to AI and have it transferred to the needs profile; lists the missing questions.
- Differentiate between must-haves/nice-to-haves: Clarify the two categories.
- Match with portfolio: Provide an anonymous portfolio list and create a short list (3-5) according to the criteria.
- Advisor filter: Check each suggestion against your own knowledge — add or remove model-agnostic factors (building management, neighborhood, actual situation).
- Suggestion message: Print a polite, understated draft of the suggestion to be sent to the recipient; no guarantee regarding price/law.
Template 1 — Needs profile from interview note:
Role: Real estate consultant assistant. Organize and organize the following anonymous interview note under the following headings: budget, financing, room/m², location preference, moving time, MUST HAVE, NICE-HAVE, motivation. List missing/unasked critical questions separately. Do not include any discriminatory criteria. Note: [anonymous note]
Template 2 — Shortlist with portfolio:
Below I have a buyer profile and an anonymous portfolio list. Simply list the 5 most suitable properties based on legitimate criteria (budget, room, m², location, financing); write "why does it fit / what criteria does it not meet" for each. DO NOT use criteria such as religion, ethnicity, marital status. Profile: [X]. Portfolio: [list]
Template 3 — Suggestion message to recipient:
Draft a warm but understated message (80-100 words) recommending these 3 properties to the buyer. Using warranty/certainty statement; Leave the price and credit to "let's discuss". Properties: [anonymous summary]
Template 4 — Discrimination/bias check:
Check my matching criteria and eliminations below for discrimination and KVKK. If there is a criterion that is illegitimate (religion, origin, marital status, disability, etc.) or violates privacy, mark it and correct it. Criteria: [list]
Comparison chart: legitimate vs prohibited criteria
criterion
Is it legitimate?
note
Budget range
✔
financial fitness
Number of rooms / m²
✔
need based
Location / district preference
✔
Buyer's own choice
Credit/financing availability
✔
solvency
Religion/ethnicity
✘
Discrimination, prohibition
Marital status / family structure
✘
Discrimination, prohibition
Disability status (for screening purposes)
✘
Forbidden; accessibility may be asked as a need
three mini cases
Case 1 — Finding the real need. The buyer wants a "detached house with a garden" but his budget is not enough. The consultant structures the needs analysis with AI: the main request is "open space for the child + proximity to the school". YZ shortlists 3+1s that fit the budget and have ample green space on site. Buyer gets one of them. Lesson: solve the underlying need, not the surface demand.
Case 2 — Bias trap. An advisor can use the pattern in historical data to tell the model “this profile wants that neighborhood” and not show some buyers to certain areas at all. This is both discriminatory and blind matching; The buyer finds the house he is looking for in another office. Lesson: you set the criteria, don't make the model a bias generator.
Case 3 — Checking the shortlist. AI recommends 5 properties that fit the criteria; But one of them has a problem that the consultant knows about but is not in the data (building management dispute, dues debt). The advisor removes it from the list. Lesson: AI recommendation shortlist, advisor filter final decision.
Weak prompt / Strong prompt
Weak prompt:
You choose the house that suits this customer and tell me which one is best.
“You choose” leaves the decision up to the model; The criteria are unclear, discrimination and error cannot be controlled.
Powerful prompt:
Shortlist only 5 properties from my portfolio that fit the following buyer profile with legitimate criteria (budget,rooms, m², location, financing); Write the reason for compliance and the shortcomings for each. Using discriminatory criteria. This is a list of suggestions, I will make the final decision. Profile: [X]. Portfolio: [list]
Common mistakes
- Elimination based on discriminatory criteria. Religion, origin, marital status — prohibited.
- Allowing the model to replicate the bias in the data.
- Solving the surface demand and skipping the real need.
- Sending the shortlist to the buyer without checking it. There may be problems that do not know the model.
- Giving customer data to the model without de-identifying it. KVKK.
Tip: Call the match “consultant shortlist,” not “system decision.” In this way, you put the responsibility both towards the customer and legally in the right place - yourself.
Buyer prioritization: warmth, not bias
The consultant's time is limited; There is a legitimate need to determine which buyer will spend their time first. But when doing this, the criterion should be "who will actually buy/rent" — not who is "sympathetic". AI can help translate buyer behavioral cues (clarified budget, loan pre-approval, specific moving date, showing up, quick turnaround) into a “readiness/warmth” score. This is legitimate because it measures readiness to buy, not the person's identity. The boundary is again clear: temperature score; It should not be based on personal characteristics such as religion, origin, marital status, age, but only on concrete steps related to the purchasing process. A buyer may be "cold" today, but will be "hot" two weeks later when their loan is approved; The score is not a static label, but an updated state. Consider the prioritization produced by AI as a suggestion: sometimes you sense a signal (a receiver's determination) that is not reflected in the data. The score guides you, it does not decide your place; and no buyer will be excluded from the list based on an illegitimate criterion.
In summary
Artificial intelligence structures customer notes, helps extract real need and produces a quick shortlist with portfolio. But matching should be based only on legitimate, property-related criteria; No discriminatory areas should be given to the model; Customer data should be de-identified. The AI's suggestion is not final, it is a shortlist that you will check and finalize.
Application task
Take a buyer call note (anonymous). Create a needs profile and a list of missing questions with template 1; Make the distinction between must-have/nice-to-have. Have Template 2 produce a shortlist of 5 from the portfolio and check each suggestion with your own knowledge. Check your criteria for discrimination/KVKK with Template 4.
checklist
- [ ] I based the matching solely on legitimate, property-related criteria.
- [ ] I did not use discriminatory criteria such as religion/origin/marital status.
- [ ] I de-identified customer data (KVKK).
- [ ] I solved the real need (must have/nice to have), not the surface demand.
- [ ] I have checked the shortlist to my own knowledge; I made the final decision.