What we build
Every layer of the lead journey — scored, explained, routed.
Each capability runs in production against your live enquiry feed — wired to your CRM, WhatsApp, and portal accounts on day one.
Context-aware BANT extraction
Budget, authority, need, and timeline pulled out of the web form, WhatsApp thread, and chat history — ready to read, not buried across five tabs.
Fraud + ID verification at source
IC and passport OCR, liveness, duplicate detection, and bot-pattern velocity checks run before the lead lands in any agent's queue.
Reasoning-based lead scoring
Scores blend BANT, intent, affordability, and source-trust into one explainable number — each backed by feature-level reasoning the agent can read.
Territory + agent-fit routing
Project, language, segment history, and past-win pattern decide the assignment. Priority leads reach a named agent in under a minute.
Intent classification · investor · own-stay · agent-rep
Separate playbooks per intent — investor enquiries flow into yield-led nurture, own-stay go to walkthrough scheduling, agent-rep handled via co-broke desk.
Disposition learning loop
Every booked · met · bought · dead outcome feeds back into the scoring model weekly — so next month's priority queue is sharper than this month's.
Who runs this
Every lead source, one priority queue.
Same scoring engine, intent classifier, and agent-fit routing — tuned to each motion. Whether you're selling a launch weekend in KL, resale in Sentosa, or international units through PropertyGuru — the stack is the same; only the training data and playbooks differ.
New-launch developer sales
High enquiry volume across launch weekend — scored, deduplicated, and routed before a single phone rings. Agents only dial high-intent, verified leads.
Landed + luxury resale
Agent-fit matching on language, past-win territory, and price band. Long-cycle leads nurtured in the background while priority queues stay clean.
International portal lead intake
PropertyGuru, iProperty, 99.co, and EdgeProp leads unified against a single profile. Duplicate submissions across portals stitched before routing.
WhatsApp-first markets
WhatsApp Business API lands in the same inbox as web forms and portals. BANT extraction reads the thread so agents don't rescroll 40 messages to catch up.
Investor + yield-led enquiries
Intent classifier flags investor vs. owner-occupier at the form level — investors routed to yield-and-exit decks, owner-occupiers to walkthrough scheduling.
Agencies with 50+ agents
Routing, supervisor override, and disposition capture in one console — with per-agent conversion dashboards so managers coach on data, not gut feel.
Model families we deploy
Four specialised models. One explainable score.
Each model owns a part of the journey — extraction, fraud, routing, feedback. Blended outputs plus per-model reasoning gives you coverage a single classifier can't match.
Fine-tuned language model reads forms, chat transcripts, and WhatsApp threads end-to-end. Returns structured BANT fields with confidence per slot.
Combines IC/passport verification, liveness match, form velocity, copy-paste signals, and duplicate history into a single fraud-risk score per submission.
Graph neural network over agent-lead interactions and closed-won patterns. Surfaces the three best-fit agents with explainable match reasoning.
Closes the loop with booked · met · bought · dead outcomes. Rebalances thresholds and agent-fit weights weekly against the latest CRM dispositions.
Data sources wired into every decision
Every signal, integrated — every decision, traceable.
Pulled in parallel, normalised to a single lead schema, versioned alongside the model consuming them.
Explainability, not just scores
A number alone doesn't earn an agent's trust. A reasoning trail does.
Every score carries the BANT extractions, intent classification, and agent-fit rationale that produced it. Agents read the trail before they dial; managers read it before they override.
- Per-lead BANT extraction with per-slot confidence
- Intent classification with source-of-evidence quote
- Agent-fit match score with feature-level explanation
- Full audit record — every score, every routing, every override
Frameworks we align to
Why Axccelerate for property lead qualification
Not a lead router.
A qualification system.
Portal tools hand you leads. Our stack qualifies them — with reasoning, fraud screening, agent-fit, and retraining every week against your closed-won pattern.
Pricing
Priced to the portfolio, not the enquiry count.
Every deployment is custom — we scope against your projects, agent pool, and CRM before quoting.
Glossary
The vocabulary behind every lead.
A reference for the terms your sales team, compliance team, and scoring model will all use.
- BANT
- Budget · Authority · Need · Timeline
The four-part qualification framework used by sales teams to decide if a lead is worth pursuing. Our extraction model reads all four out of form and chat history — with per-field confidence scores attached.
- SPIN
- Situation · Problem · Implication · Need
A consultative qualification framework — more thorough than BANT for high-value residential. Useful for landed and luxury resale where motivations are compound and timeline is elastic.
- Lead score
- Ranked enquiry priority
A 0-100 number combining BANT, intent, affordability, and source-trust. 70+ routes to priority, 40-69 to warm nurture, below 40 to low-priority with no human contact until the lead warms.
- Dead lead
- Disposition · unreachable or disqualified
A lead marked dead after a set number of outreach attempts without response, or after explicit disqualification. Dead dispositions still feed the model — they sharpen future scoring.
- MQL
- Marketing Qualified Lead
A lead that has demonstrated enough engagement (multiple page visits, brochure download, return visit) to warrant sales outreach — but hasn't yet passed BANT screening.
- SQL
- Sales Qualified Lead
A lead that has passed BANT screening and is ready for direct agent engagement. Priority queue members are SQLs by definition in most of our deployments.
- Disposition
- Outcome of a lead
The terminal state — booked viewing, met, signed booking fee, purchased, dead, on-hold. Every disposition is a training signal for the feedback learner.
- Pre-approval
- Loan pre-approval
A lender's conditional indication of how much a buyer can borrow, based on income and credit checks. Pre-approved leads score materially higher on the authority dimension.
- TDSR
- Total Debt Servicing Ratio
The Singapore regulatory cap on what proportion of income can go to debt repayments. Also a useful affordability proxy in Malaysia; used in our own-stay scoring path.
- ABSD
- Additional Buyer's Stamp Duty
Singapore's tax on second and subsequent residential purchases. Signals investor intent and materially shifts scoring thresholds for foreign and second-home buyers.
- LTV
- Loan-to-Value ratio
Loan amount divided by property value. Mortgage caps live here — LTV limits shape what price brackets a buyer can actually hit, which feeds affordability scoring.
- Owner-occupier vs investor
- Intent segmentation
The single most important classification on any enquiry. Separate playbooks — own-stay goes to walkthroughs, investor goes to yield decks — drive different agent matches and nurture paths.
- KYC / AML
- Know Your Customer / Anti-Money-Laundering
Identity verification and source-of-funds screening. Mandatory for developer sales above thresholds in SG, MY, ID — runs automatically at enquiry ingest, not at contract.
- Cooling-off period
- Statutory reconsideration window
The period during which a buyer can withdraw without penalty after signing. Scoring doesn't celebrate a conversion until this window passes — early churn is a real signal.
Turn every enquiry into a scored, reasoned priority.
30-minute scoping with a senior engineer and a lead-systems operator. You'll leave with a scoring plan, integration sketch, and realistic timeline — not a sales pitch.