Intent Recognition is the first intelligence layer that converts unstructured prospect, family, and referrer communication into structured CRM signals. It helps Journey understand what a person is asking, how urgent it is, which record it belongs to, and what the system should recommend next.
In Phase 1, Journey CRM captures and organizes communications. In Phase 2, the Ghost Assistant begins to interpret those communications. Intent Recognition reads the content of inbound messages, determines the likely business meaning, assigns confidence, and produces a structured signal that downstream features can use.
Understand
Detect whether the message is about pricing, tour scheduling, availability, care needs, complaints, pet policy, referrer activity, or follow-up.
Prioritize
Use urgency and engagement signals to support lead scoring, morning brief, hot lead alerts, and manager visibility.
Assist
Trigger the correct assistive action: response draft, task suggestion, appointment prompt, tag update, or human review.
Core idea: Intent Recognition is not just message labeling. It is the decision layer that helps Journey CRM decide what should happen next.
2. Primary Users and Personas
Leasing Agent
Needs to quickly understand what a lead wants and respond without reading every prior message.
Sales Director
Needs to identify urgent/high-intent leads and make sure follow-ups are not missed.
Executive / Owner
Needs to see intent trends across properties, agents, and lead sources.
Operations Admin
Needs AI classifications to be configurable, auditable, and safe for customer adoption.
3. Planned Intent Taxonomy
The taxonomy should start practical and expand as Journey receives more real-world communication data. Each detected intent should include a confidence score and optional supporting reason.
Intent
Example Message
Recommended CRM Action
Pricing Inquiry
“What is the monthly cost?”
Draft pricing response, raise score, tag pricing interest.
Route for human review and avoid unsupported automation.
4. Workflow Diagram
The ideal workflow keeps AI interpretation separate from final action. This lets Journey add intelligence without forcing every customer into fully autonomous automation.
flowchart TD
A[Inbound Communication] --> B{Source}
B -->|SMS| C[Normalize Message]
B -->|Email| C
B -->|Chat / Widget| C
B -->|Call Transcript| C
C --> D[Attach CRM Context]
D --> E[Intent Recognition Model]
E --> F{Confidence Level}
F -->|High| G[Create Intent Signal]
F -->|Medium| H[Suggest Intent for Review]
F -->|Low| I[Route to Human Review]
G --> J[Downstream Actions]
H --> J
J --> K[Draft Response]
J --> L[Suggest Task]
J --> M[Update Lead Score]
J --> N[Morning Brief / Alerts]
J --> O[Reporting & Dashboard]
5. Detailed User Cases
After-hours inquiryPricingPet policy
User Case 1: Family asks about pets and pricing at 2:00 AM
Scenario: A daughter texts the community after hours: “Do you allow small dogs and what is the monthly pet fee?”
Current friction: Without intent recognition, the lead may receive a generic response and wait until morning for a useful answer.
System behavior: AI classifies the message as Pet Policy + Pricing Inquiry, checks property-level content if available, drafts an agent-safe response, and flags the lead as engaged.
Result: The agent begins the next day with context, a prepared draft, and a higher-priority lead.
UrgencyDischargeManager alert
User Case 2: Hospital discharge creates urgent move-in need
Scenario: A family member writes: “My father is being discharged Friday. We need a place quickly. Can someone call me today?”
System behavior: AI detects Urgent Move-in / Discharge, assigns high urgency, raises lead score, suggests immediate follow-up, and can surface the lead in Morning Brief or manager alerts.
Result: The lead does not sit behind lower-priority inquiries. The team acts before the opportunity is lost.
ComplaintEscalationHuman review
User Case 3: Lead expresses frustration
Scenario: A prospect says: “I already filled this out twice and nobody called me back.”
System behavior: AI detects Complaint / Frustration, avoids automatic customer-facing action, creates an escalation suggestion, and prepares a careful apology/follow-up draft for review.
Result: The team can recover the relationship while keeping sensitive customer communication under human control.
ReferrerProfessional sourcePipeline creation
User Case 4: Referrer sends a potential resident
Scenario: A discharge planner emails: “I have a patient who needs assisted living and the family wants options near your property.”
System behavior: AI detects Referrer Inquiry, identifies professional referral context, suggests creating or linking a referrer record, and prepares a prospect intake task.
Result: Referral partner activity becomes more structured and easier to measure.
6. Proposed User Experience
In Communication Thread
Show detected intent badges near the message: Pricing, Tour Request, Urgent, Complaint, etc. Include confidence and “Review” for uncertain classification.
On Guest Card
Show latest intent, recent intent history, and recommended actions: draft reply, add task, schedule appointment, or escalate.
On Prospect List
Add intent-based filters and badges so agents can quickly find pricing inquiries, tour requests, urgent leads, and no-response risks.
In Manager Dashboard
Aggregate intent trends by property, source, agent, and referrer to understand what prospects are asking and where conversion opportunities exist.
7. Data, Events, and CRM Integration
Intent Recognition should create a structured output that downstream modules can use without re-running the AI model every time.
Field
Purpose
intent_type
Primary classification such as pricing_inquiry, tour_request, urgent_move_in.
secondary_intents
Additional detected intents such as pet_policy + pricing.
confidence_score
Numeric confidence for review and automation gating.
urgency_level
Low, medium, high, urgent.
source_channel
SMS, email, chat, widget, call transcript.
guest_card_id
Linked CRM record.
recommended_action
Draft reply, task suggestion, schedule appointment, manager review, no action.
