hasbriSystem logic
Built to be inspectable

Every feature.
The logic behind it.

What Hasbri listens to, what it calculates, what it will not assume, and when a human stays in control.

AI interpretationUnderstands language using the existing analysis cycle.
Deterministic ruleRepeatable calculation or safety guardrail.
Live evidenceCurrent transcript, session, or public listing data.
Human authorityManager approval is required for commitments.
01Conversation input

Live transcription

Microphone audio becomes a speaker-separated transcript. Speaker identity is inferred from conversational evidence; uncertain identity stays Unknown.

MicrophoneTranscriptSpeaker roles
Evidence rule

Does not assume the first speaker is the customer or salesperson.

02AI interpretation

Customer understanding

Extracts needs, budget, trade-in, timeline, vehicle preferences, objections, and buying signals only from statements attributed to the customer.

Customer speechGrounded factsProfile
Evidence rule

Unknown information remains blank. Salesperson suggestions are not converted into customer facts.

03AI + guardrails

Decision support

Surfaces one next move after a material priority, objection, buying signal, inventory decision, or stage change. Otherwise it stays quiet.

Material changeEvidence checkOne next move
Evidence rule

Never invents an authorized price or displays constant coaching.

04Salesperson coaching

Conversation Health

Reuses the existing conversation analysis and combines it with objective transcript metrics. It evaluates the salesperson—not the customer.

Seller-attributed speechAI + rulesQuiet coaching
Evidence rule

Unknown speaker roles show Unknown. Seller profanity, long monologues, missed discovery, and ignored signals can trigger one coaching message.

05Deterministic ranking

Inventory recommendation

Ranks real dealership listings against hard requirements and softer preferences. Aesthetic words such as pretty or sleek never eliminate a vehicle.

Customer requirementsDealer inventoryRanked match
Evidence rule

A single $30,000 statement searches a flexible $20,000–$40,000 range. Exact year, make, model, or stock requests remain hard constraints.

06Verified comparables

Market price

Uses verified comparable listings when enough evidence exists. It separates the dealer listing price from market context.

Selected vehicleComparable listingsMarket range
Evidence rule

If comparable evidence is insufficient, the feature reports unavailable instead of estimating.

07Operational intelligence

Manager Live Floor

Shows active conversations, stage, customer signal, recommended manager action, walking risk, and pending approvals across the floor.

Live sessionsPriority signalsManager action
Evidence rule

Rows use current session data. Inactive sessions fall off the live view after the heartbeat expires.

08Human authority

Pricing & approvals

Sends the manager the exact selected vehicle, listed price, customer request, buying intent, and transcript-supported signals without salesperson retyping.

Selected listingApproval requestManager response
Evidence rule

The AI may explain a request, but only the manager authorizes price or deal terms.

09Market intelligence

Demand Radar

Combines dealership-scoped public inventory with discussion signals. Brand dealerships only see their brand; general dealerships can cover all makes.

Public inventoryDemand signalsDealer action
Evidence rule

Live inventory and sample discussion data are labeled separately. Sample data is never presented as live.

Honesty by design

What the platform does not pretend to know.

Credit use

One analysis cycle, several useful outputs.

Customer Understanding, Conversation Health, stage, intent, and summary reuse the same structured conversation analysis. Conversation Health does not make a separate AI request. Inventory is narrowed before AI guidance receives vehicle context.