Skip to content
Trilogy Care Logo TC Docs

AI Automation Initiatives - TC Portal

AI Automation Initiatives - TC Portal

Comprehensive roadmap of AI/ML automation opportunities across the Trilogy Care workflow lifecycle

Overview

This document outlines automation opportunities across 6 key stages of the care management process. Each initiative aims to reduce manual work, improve accuracy, and enable care teams to focus on high-value client interactions.

Total Estimated Monthly Impact: ~14,000-17,000 automated records/tasks

Stage 1: Lead Management

1.1 Lead Calls (Sales Phone Only)

Current Process: Sales team manually takes notes during calls with prospective clients.

Automation Opportunities:

  • AI transcriptions to summarize notes, score call intent, identify lead traits, and prioritize next steps
  • Extract contact details, categorize intent, and flag risks
  • Generate picklists to capture what’s most important to the lead (e.g., services, location, cost)

Estimated Impact: ~1,000-2,000 records/month

Key Benefits:

  • Faster lead qualification
  • Consistent data capture
  • Automated lead scoring and prioritization

Target Users:

  • Sales Team (STEPHEN, JACQUI, BERNIE)
  • Onboarding Team

Technical Considerations:

  • Voice transcription API (e.g., AssemblyAI, Deepgram, OpenAI Whisper)
  • NLP for intent classification and entity extraction
  • Integration with CRM/lead management system

1.2 Questionnaire

Current Process: Manual data entry from questionnaire forms and government portal PDFs (ACAT/IAT).

Automation Opportunities:

  • Extract questionnaire data from form submissions
  • Scrape government portal for PDF ACAT/IAT data, convert to JSON
  • Use AI to categorize recipient data and generate recommended care plan, budget, and risk assessment

Estimated Impact: ~400-500 records/month

Key Benefits:

  • Eliminate manual data entry
  • Auto-generate initial care plan drafts
  • Identify risks early in the process

Target Users:

  • Assessment Team (PAT & SIAN)
  • Care Partners

Technical Considerations:

  • PDF parsing (e.g., PyPDF2, pdfplumber, or OCR for scanned docs)
  • Form data extraction and validation
  • AI model for care plan generation (LLM with prompt engineering)
  • Integration with Package/BudgetPlan domains

Stage 2: Needs Assessment

2.1 Medical Summary

Current Process: Manual review of Medical Health Summary PDFs to identify risks and needs.

Automation Opportunities:

  • Extract text from Medical Health Summary assessment PDFs
  • Use AI to cross-reference with other assessment data (questionnaire, care partner notes) to identify risks, needs, and goals
  • Summarize data into actionable items

Estimated Impact: ~200-300 records/month

Key Benefits:

  • Faster risk identification
  • Comprehensive cross-referencing of multiple data sources
  • Actionable summaries for care teams

Target Users:

  • Assessment Team
  • Clinical Team (Clinical Nurse)
  • Care Partners

Technical Considerations:

  • PDF text extraction
  • AI summarization and cross-referencing
  • Risk scoring/classification model
  • Integration with clinical records and Package needs

2.2 Meeting Conducted (Post-Assessment)

Current Process: Manual note-taking and care plan refinement after assessment meetings.

Automation Opportunities:

  • Summarize post-assessment meeting transcripts using AI to refine care plans and suggest additional services
  • Provide recommendations for care plan improvement
  • Compare notes with initial care plan and generate updated content

Estimated Impact: ~400-500 records/month

Key Benefits:

  • Consistent meeting documentation
  • AI-suggested care plan improvements
  • Version tracking of care plan changes

Target Users:

  • Assessment Team
  • Care Partners
  • Clinical Team

Technical Considerations:

  • Meeting transcription (Zoom API, AssemblyAI, etc.)
  • AI comparison of initial vs. updated care plans
  • Care plan versioning (potentially event sourcing)
  • Integration with BudgetPlan domain

Stage 3: Care Planning

3.1 Care Plan Written

Current Process: Care Partners manually write care plans; manual QA for compliance.

Automation Opportunities:

  • AI proofing for consistency, grammar, spelling errors, and formatting compliance
  • Ensure care plan aligns with regulatory standards (e.g., measurable goals, review schedules)
  • With holistic care plan, identify vulnerability rating and other high-level classifications (tags) for the Package

Estimated Impact: ~400-500 records/month

Key Benefits:

  • Regulatory compliance assurance
  • Consistent formatting and quality
  • Auto-classification of vulnerability and risk levels

Target Users:

  • Care Partners
  • Quality Team (ERIN, MEAGAN)

Technical Considerations:

  • AI grammar/compliance checker (GPT-4, Claude)
  • Regulatory rules engine
  • Tagging and classification system for Packages
  • Integration with Package domain

3.2 Assigning a Case Manager

Current Process: Manual assignment of Care Partners/Coordinators based on location, language, and client needs.

Automation Opportunities:

  • Look up internal database to find the most suitable Care Partner and/or Coordinator based on:
    • Geographic location
    • Language preferences
    • Client needs
    • Workload capacity
  • AI recommends best-fit assignments with reasoning for the choice

Estimated Impact: ~400-500 assignments/month

Key Benefits:

  • Faster, data-driven assignments
  • Better care partner/client matching
  • Balanced workload distribution

Target Users:

  • Case Managers (supervise coordinators)
  • Administrators
  • Onboarding Team

Technical Considerations:

  • Requires storing detailed Care Partner profiles (location, languages, specialties, capacity)
  • Matching algorithm (rule-based + ML scoring)
  • Integration with User domain and Package assignments

Dependencies:

  • Need to build Care Partner profile management system first

Stage 4: Care Delivery (Calls Made)

4.1 General Call Classifications

Current Process: Manual note-taking and task creation after calls with clients.

Automation Opportunities:

  • AI transcription and summarization of calls
  • Generate general tasks and action items for the right stakeholders
  • Custom tagging: Apply Trilogy’s custom tag classifications (for Notes)
  • Log summary + transcript + tags to call records in CRM/Portal for easy call reviewing

Estimated Impact: ~10,000 calls/month

Key Benefits:

  • Automatic call documentation
  • Consistent tagging and classification
  • Actionable task generation
  • Searchable call history

Target Users:

  • Care Partners
  • Care Coordinators
  • Clinical Team

Technical Considerations:

  • Real-time or batch transcription
  • AI summarization and task extraction
  • Custom tag taxonomy and classification model
  • Storage: summaries in database, full transcripts in blob storage (S3/similar)
  • Integration with Notes/Tasks system

Privacy/Compliance:

  • GDPR/privacy compliance for call recordings
  • Secure storage and access controls
  • Consent and notification requirements

4.2 Need/Goal/Budget Change

Current Process: Manual detection of care plan changes discussed in calls/emails.

Automation Opportunities:

  • Analyze calls and emails for discussions related to changes in:
    • Client needs
    • Living situation
    • Goals
  • Trigger workflows to update care plans, budgets, or service utilization in draft stage
  • Require Care Partner approval before committing changes

Estimated Impact: TBD (subset of 10,000 calls)

Key Benefits:

  • Proactive care plan updates
  • Reduced risk of outdated care plans
  • Automated draft creation with human-in-the-loop

Target Users:

  • Care Partners
  • Care Coordinators

Technical Considerations:

  • NLP model to detect change signals
  • Draft budget/care plan creation
  • Workflow/approval system
  • Integration with BudgetPlan and Package domains
  • Event sourcing for audit trail

4.3 Incidents Logged

Current Process: Manual incident logging after calls.

Automation Opportunities:

  • Log specific incidents (e.g., falls, medication issues) flagged during calls
  • Categorize incidents by severity and trigger workflows for follow-up or escalation
  • Create recommendations and action plans that fit clinical policy

Estimated Impact: ~300-500 incidents/month

Key Benefits:

  • Faster incident reporting
  • Consistent categorization and severity scoring
  • Automated escalation workflows

Target Users:

  • Care Partners
  • Clinical Team
  • Quality Team

Technical Considerations:

  • Incident classification model
  • Severity scoring algorithm
  • Escalation workflow engine
  • Integration with incident management system (if exists) or new Incident domain
  • Compliance with Serious Incident Response Scheme

4.4 Risks Identified

Current Process: Manual risk identification during calls.

Automation Opportunities:

  • Detect minor risks based on Risk Criteria during calls (keywords/phrases like fall risks, health changes)
  • Assign severity scores and log risks in risk management system

Estimated Impact: ~100-200 risks/month

Key Benefits:

  • Proactive risk detection
  • Consistent risk scoring
  • Centralized risk tracking

Target Users:

  • Care Partners
  • Clinical Team

Technical Considerations:

  • Risk keyword/phrase detection model
  • Risk severity scoring
  • Integration with Package risk management
  • Relationship to clinical team workflows

4.5 Refine Risk (Backlog)

Current Status: Backlog item - current risk data is not refined.

Automation Opportunities:

  • Work through backlog of existing risk data in context with other client data
  • When risks are identified by Care Partner, classify and refine shorthand info
  • Standardize risk data format

Estimated Impact: TBD (one-time backlog processing + ongoing)

Key Benefits:

  • Clean, structured risk data
  • Better reporting and analytics
  • Improved clinical decision-making

Target Users:

  • Clinical Team
  • Care Partners
  • Data/Analytics Team

Technical Considerations:

  • Data migration and cleanup
  • Risk taxonomy and standardization
  • AI-assisted classification of existing unstructured data

4.6 Feedback

Current Process: Manual feedback review and classification.

Automation Opportunities:

  • Analyze feedback from clients or families to help Operations team classify it
  • Identify potential feedback or complaints made via phone calls
  • Flag sentiment (positive/negative) on a scale of 1-5
  • Generate reports for service improvement

Estimated Impact: ~100-200 feedback items/month

Key Benefits:

  • Automated sentiment analysis
  • Faster identification of complaints
  • Data-driven service improvement

Target Users:

  • Quality Team
  • Operations Team
  • Care Partners

Technical Considerations:

  • Sentiment analysis model
  • Feedback classification taxonomy
  • Reporting dashboard
  • Integration with Quality domain

4.7 Requests (Cab Charge, OT Reports, Meals, etc.)

Current Process: Manual detection and processing of client requests.

Automation Opportunities:

  • Detect specific client requests during calls (e.g., cab charge, OT reports, meals)
  • Validate against client’s budget
  • Recommend budget adjustments or create tasks
  • Ensure budget utilization aligns with service requirements

Estimated Impact: ~100-200 requests/month

Key Benefits:

  • Faster request processing
  • Automatic budget validation
  • Task creation for fulfillment

Target Users:

  • Care Partners
  • Finance Team
  • Service Providers

Technical Considerations:

  • Request entity extraction
  • Budget validation logic
  • Task/workflow creation
  • Integration with BudgetPlan and Billing domains

4.8 Package Upgrade

Current Process: Manual detection and processing of package upgrade requests.

Automation Opportunities:

  • Detect Package Upgrade Requests from calls (different logic from general tasks)
  • Automate workflows to notify care teams about package changes
  • Trigger reassessment or budget review processes

Estimated Impact: ~50-100 requests/month

Key Benefits:

  • Faster package upgrade processing
  • Automated notification workflows
  • Better tracking of upgrade requests

Target Users:

  • Care Partners
  • Sales/Onboarding Team
  • Finance Team

Technical Considerations:

  • Specialized request detection (different from general requests)
  • Workflow automation for upgrades
  • Integration with Package domain
  • Possible integration with Services Australia API

4.9 Check-ins

Current Process: Manual documentation of routine check-in calls.

Automation Opportunities:

  • Extension of General Call Summary
  • Tag phone calls as “Check-in” and summarize notes
  • Use AI to contextualize unique note type to write check-in notes reflective of recent goals, needs, and budgets
  • Generate actionable insights for care teams and families

Estimated Impact: ~200-400 check-ins/month

Key Benefits:

  • Consistent check-in documentation
  • Context-aware note generation
  • Actionable insights for ongoing care management

Target Users:

  • Care Partners
  • Care Coordinators

Technical Considerations:

  • Check-in classification
  • Context-aware summarization (pull in recent care plan, goals, budgets)
  • Integration with Package and BudgetPlan data

Stage 5: Bill Management

5.1 Bill Analysis

Current Process: Manual extraction and categorization of bill data.

Automation Opportunities:

  • Extract data from bills (e.g., client name, provider, line items) into JSON format
  • Categorize line items and match against client budgets
  • Recommend adjustments or highlight discrepancies using match suitability formula for budget alignment
  • Automate insights into budget utilization

Estimated Impact: ~200-400 bills/month

Key Benefits:

  • Faster bill processing
  • Automatic budget matching
  • Early discrepancy detection
  • Better budget utilization insights

Target Users:

  • Finance Team (MELLETTE, GUS)
  • Care Partners

Technical Considerations:

  • PDF/document parsing
  • Line item classification
  • Budget matching algorithm
  • Integration with Billing and BudgetPlan domains
  • Potential integration with existing bill processing workflows

5.2 Bill Notes for Case Management

Current Process: Manual review of bill notes to identify case management implications.

Automation Opportunities:

  • Review bill notes and use AI to suggest tasks or next steps for case managers
  • Flag unusual expenses
  • Recommend budget reviews
  • Feed insights back into care plans or risk tracking systems

Estimated Impact: ~100-200 bill notes/month

Key Benefits:

  • Proactive case management insights from financial data
  • Early detection of budget issues
  • Better integration of financial and care planning

Target Users:

  • Care Partners
  • Finance Team

Technical Considerations:

  • Bill note analysis and task generation
  • Integration with case management workflows
  • Feedback loop to BudgetPlan and Package domains

Stage 6: OT Reports / Inclusions

6.1 OT Report Provided

Current Process: Manual review of Occupational Therapist reports.

Automation Opportunities:

  • Convert OT reports into text (PDF parsing)
  • Analyze reports for risks, goals, needs, and budget implications
  • Cross-reference with existing care plans and trigger updates where necessary

Estimated Impact: ~50-100 reports/month

Key Benefits:

  • Faster OT report processing
  • Automatic care plan updates
  • Risk identification from clinical reports

Target Users:

  • Clinical Team
  • Care Partners
  • Assessment Team

Technical Considerations:

  • PDF text extraction
  • Clinical entity extraction (risks, goals, recommendations)
  • Care plan update workflows
  • Integration with Package and clinical data

6.2 Inclusion/Exclusion

Current Process: Manual processing of inclusion/exclusion requests based on 11-question framework.

Automation Opportunities:

  • Automate identification of inclusion/exclusion requests
  • Apply 11-question framework for decision support
  • Prefill relevant fields in records for faster processing

Estimated Impact: ~50-100 requests/month

Key Benefits:

  • Faster inclusion/exclusion decisions
  • Consistent application of framework
  • Reduced manual data entry

Target Users:

  • Care Partners
  • Clinical Team
  • Finance Team

Technical Considerations:

  • 11-question framework logic (decision tree or rule engine)
  • Request classification
  • Form prefilling
  • Integration with service approval workflows

Technical Architecture Considerations

Core AI/ML Infrastructure

Transcription & Speech-to-Text:

  • AssemblyAI, Deepgram, or OpenAI Whisper
  • Real-time vs. batch processing
  • Language support (multilingual clients)

Natural Language Processing:

  • OpenAI GPT-4 / Claude for summarization, classification, task extraction
  • Fine-tuning for domain-specific terminology (aged care, HCP)
  • Prompt engineering and template management

Document Processing:

  • PDF parsing: PyPDF2, pdfplumber, or Tesseract OCR for scanned docs
  • Form data extraction
  • Structured data output (JSON)

Data Storage:

  • Summaries and structured data: PostgreSQL (existing)
  • Full transcripts and large documents: Blob storage (S3, Azure Blob)
  • Vector embeddings for semantic search: Pinecone, Weaviate, or pgvector

Queue and Async Processing:

  • Laravel Queues + Horizon (existing)
  • Background jobs for AI processing
  • Webhook handlers for external API callbacks

Monitoring & Logging:

  • AI job success/failure rates
  • Processing times and costs
  • Accuracy metrics (human-in-the-loop feedback)

Privacy, Security, and Compliance

Key Considerations:

  • Consent: Client consent for call recording and AI processing
  • Data Privacy: Compliance with Australian Privacy Principles, GDPR
  • Secure Storage: Encrypted storage for sensitive data
  • Access Controls: Role-based access to AI-generated content
  • Audit Trail: Event sourcing for AI decisions and human overrides
  • De-identification: Consider de-identifying data for model training

Serious Incident Response Scheme:

  • Ensure AI-flagged incidents comply with mandatory reporting requirements
  • Clinical team review and approval before submission

Implementation Priorities

Phase 1: High-Impact, Low-Complexity

  1. General Call Classifications (Stage 4.1) - Highest volume, foundational
  2. Care Plan Proofing (Stage 3.1) - Immediate quality improvement
  3. Bill Analysis (Stage 5.1) - High ROI, existing pain point

Phase 2: Risk and Compliance

  1. Incidents Logged (Stage 4.3) - Compliance and safety
  2. Risks Identified (Stage 4.4) - Proactive risk management
  3. Medical Summary (Stage 2.1) - Clinical decision support

Phase 3: Workflow Automation

  1. Need/Goal/Budget Change (Stage 4.2) - Proactive care plan updates
  2. Lead Calls (Stage 1.1) - Sales pipeline optimization
  3. Assigning a Case Manager (Stage 3.2) - Workload balancing

Phase 4: Advanced Features

  1. Questionnaire (Stage 1.2) - End-to-end onboarding automation
  2. OT Reports (Stage 6.1) - Clinical integration
  3. Refine Risk Backlog (Stage 4.5) - Data quality improvement

Success Metrics

Efficiency Metrics:

  • Time saved per Care Partner/Coordinator (hours/week)
  • Reduction in manual data entry (%)
  • Faster turnaround times (lead-to-client, incident-to-resolution)

Quality Metrics:

  • Care plan compliance rate
  • Incident detection accuracy
  • Risk identification recall/precision

Financial Metrics:

  • Cost savings from automation
  • AI processing costs vs. manual labor costs
  • Budget utilization improvement

User Satisfaction:

  • Care Partner satisfaction with AI tools
  • Reduction in administrative burden (survey)
  • Client satisfaction with care responsiveness

Next Steps

  1. Prioritize specific initiatives based on business impact and technical feasibility
  2. Prototype high-priority features (e.g., call transcription + summarization)
  3. Pilot with small user group (e.g., 5-10 Care Partners)
  4. Iterate based on feedback and metrics
  5. Scale to all users with feature flags

This document serves as a living roadmap for AI automation initiatives. It should be updated as initiatives are implemented, deprioritized, or new opportunities emerge.