MARKET LEAD

AI Implementation Plan

Incorrect password
Press Enter to continue
AI Implementation Plan

AI-Powered Client Communications for Lucerne Private Wealth

A comprehensive plan for implementing Claude AI to automate meeting follow-ups, client emails, and action plans, with full PII protection and regulatory compliance for Australian financial services.

Prepared by Market Lead March 2026 Confidential

1. Executive Summary

What this solution does and why it works for a private wealth firm

Lucerne Private Wealth wants to use AI to transform how advisors handle post-meeting client communications. Currently, after a client meeting, an advisor must manually write up notes, draft a follow-up email summarising the discussion, outline action items, and flag anything requiring client approval. This is time-consuming and inconsistent.

The proposed solution connects to Lucerne's existing Microsoft 365 environment (Outlook, Teams, Dynamics CRM) via secure APIs, processes meeting transcripts and emails through a PII removal layer, sends the sanitised content to Claude AI for intelligent summarisation and action plan generation, then re-inserts client details only at the final human-approved output stage.

Zero
PII Sent to AI
100%
Human Approval Before Send
0 Days
API Data Retention by Anthropic
Full
Australian Regulatory Alignment
The Core Principle: Strip, Process, Rehydrate

Client PII is removed before any data leaves the secure environment. Claude AI only ever sees anonymised placeholders (e.g., "[CLIENT_1]", "[ACCOUNT_REF_A]"). After AI processing, the system re-inserts real names and details in a controlled, auditable step that requires human approval before any communication is sent.

2. System Architecture Overview

How the components connect, from Microsoft 365 to Claude AI and back

LUCERNE SECURE ENVIRONMENT PII PROCESSING LAYER AI PROCESSING (NO PII) Microsoft Outlook Emails & Calendar Microsoft Teams Meetings & Transcripts Dynamics 365 CRM Client Records Microsoft Graph API PII Detection Engine NER + Regex + Dictionary TFN, ABN, Names, Accounts Tokenisation Layer Replace PII with placeholders Secure Mapping Store Token-to-PII lookup (encrypted) Rehydration Engine Re-insert real PII into output HUMAN APPROVAL GATE Claude API Zero-retention processing Anonymised data only Meeting Summariser Key points & themes Action Plan Generator Tasks, approvals, next steps Email Draft Writer Professional client follow-up NO PII AI output lookup
Key Architectural Decision

The PII Processing Layer runs entirely within Lucerne's secure environment (Azure tenant or on-premises). No personally identifiable information ever crosses the network boundary to Anthropic's servers. The AI only processes text containing generic placeholders like [CLIENT_1], [ACCOUNT_A], [ADVISER_1].

3. Microsoft Ecosystem Integration

Connecting to Outlook, Teams, and Dynamics 365 via Microsoft Graph API

Authentication: Azure AD / Entra ID (OAuth 2.0)

All API access uses Microsoft Entra ID (formerly Azure Active Directory) with OAuth 2.0 authorization code flow. This ensures:

OAuth 2.0 Authorization Flow 1. Advisor Login MFA + Device Check 2. Azure AD Validates + Issues Token 3. Scoped Token Mail.Read, Calendar.Read 4. Graph API Call Fetch emails/meetings 5. Data Returned Required API Scopes (Minimum Privilege) Outlook Emails: Mail.Read, Mail.Send (delegated per-user) Calendar: Calendars.Read (delegated per-user) Teams Transcripts: OnlineMeetingTranscript.Read.All (application, admin-consented) Dynamics CRM: Dynamics 365 Web API via separate OAuth app registration

Key API Endpoints

Outlook Emails

GET /me/messages

GET /me/messages/{id}

POST /me/sendMail

Read advisor's inbox, filter by client, draft and send follow-ups via Microsoft Graph API v1.0.

Calendar & Meetings

GET /me/events

GET /me/onlineMeetings

Retrieve upcoming and past meeting details, attendees, and links to Teams recordings.

Teams Transcripts

GET /me/onlineMeetings/{id}/transcripts

GET .../transcripts/{id}/content

Download auto-generated meeting transcripts in VTT or text format for AI processing.

Dynamics 365 CRM

GET /api/data/v9.2/contacts

PATCH /api/data/v9.2/contacts({id})

Read client records, update notes, log activities. Uses OData v4 REST API with separate Azure app registration.

Important: Teams Transcript Availability

Teams meeting transcripts require Teams Premium or Microsoft 365 E5 licensing for automatic transcription. If Lucerne is on a lower tier, transcripts can be generated via a third-party service or manual upload. Confirm current licensing before build.

4. PII Detection & Removal Engine

How personally identifiable information is stripped before AI processing

The PII engine is the security backbone of the entire system. It runs as a server-side process within Lucerne's secure environment (their Azure tenant or on-prem servers). No data leaves until PII is removed.

PII Detection & Tokenisation Pipeline Raw Input Emails, transcripts, CRM notes with PII Layer 1: Regex Patterns TFN: \d{3}\s?\d{3}\s?\d{3} ABN: \d{2}\s?\d{3}\s?\d{3}\s?\d{3} Phone: 04\d{2}[\s-]?\d{3}[\s-]?\d{3} Email: [a-zA-Z0-9+._-]+@... Layer 2: NER Model Person names Organisation names Addresses / locations Financial entities Layer 3: Dictionary Known client names Account numbers Portfolio references Custom entity list Tokenisation: Replace with Placeholders "John Smith" becomes [CLIENT_1] | "0412 345 678" becomes [PHONE_1] Encrypted Mapping [CLIENT_1] = "John Smith" Sanitised Output "[CLIENT_1] discussed their portfolio with [ADVISER_1]. They want to reallocate [AMOUNT_1] from bonds to equities." SAFE TO SEND TO AI

What Gets Detected & Replaced

PII CategoryDetection MethodExample InputTokenised Output
Client NamesNER + CRM DictionaryJohn & Sarah Smith[CLIENT_1] & [CLIENT_2]
Tax File NumbersRegex (9 digits)123 456 789[TFN_1]
ABN / ACNRegex (11/9 digits)51 824 753 556[ABN_1]
Phone NumbersRegex (AU patterns)0412 345 678[PHONE_1]
Email AddressesRegexjohn@email.com[EMAIL_1]
Physical AddressesNER + Pattern42 Collins St, Melbourne[ADDRESS_1]
Account NumbersCRM DictionaryACC-LPW-00847[ACCOUNT_1]
Dollar AmountsRegex + Context$2.4M portfolio[AMOUNT_1] portfolio
Dates of BirthRegex + Contextborn 15/03/1965born [DOB_1]
Medicare NumbersRegex (10 digits)2123 45670 1[MEDICARE_1]
Bank BSB/AccountRegexBSB 062-000, Acc 1234567BSB [BSB_1], Acc [BANKACCT_1]
Defence in Depth: Why Three Layers?

Regex alone misses context-dependent PII (e.g., "Mr. Smith" without a number pattern). NER alone misses structured data (TFNs, account numbers). Dictionary lookup alone misses new or ad-hoc mentions. The three layers together provide comprehensive coverage. A post-processing scan validates that no known PII patterns remain in the sanitised output before it leaves the secure zone.

5. End-to-End Workflow

From client meeting to approved follow-up email, step by step

Meeting-to-Follow-Up Workflow 1 Client Meeting Occurs (Teams or In-Person) Teams auto-generates transcript. For in-person, advisor uploads recording or manual notes. 2 System Fetches Data via Microsoft Graph API Pulls: meeting transcript, related emails (last 30 days with this client), CRM client profile, previous action items. 3 PII Detection & Tokenisation All client names, account numbers, TFNs, addresses, and financial details replaced with safe tokens. Mapping stored encrypted. 4 Claude AI Processes Sanitised Content Generates: meeting summary, action items with owners, items needing client approval, and draft follow-up email. All using [TOKEN] placeholders. 5 Rehydration: Real PII Re-Inserted Tokens replaced with actual names and details from the encrypted mapping store. Output now contains real client information. 6 HUMAN REVIEW & APPROVAL (MANDATORY) Advisor reviews the complete email in their Outlook drafts. Can edit any content. Must click "Approve & Send". No email is ever sent automatically. Every output requires explicit human confirmation. 7 Email Sent via Outlook Sent from advisor's actual email address 8 CRM Updated Action items logged to Dynamics 365 Audit Log: Full trail of what was processed, what AI generated, what human approved, and what was sent

Example Output: Client Follow-Up Email

DRAFT (Pending Advisor Approval)

To: john.smith@email.com
Subject: Summary of Our Meeting, 18 March 2026


Hi John and Sarah,

Thank you for meeting with us today. Here is a summary of what we discussed and the next steps:

Key Discussion Points:

  • Reviewed your current portfolio allocation (60% equities / 30% bonds / 10% alternatives)
  • Discussed increasing equities exposure to 70% given your 15+ year investment horizon
  • Reviewed the performance of the international equities allocation
  • Discussed updating your estate planning documents following recent legislative changes

Action Items:

  • We will: Prepare a rebalancing proposal for the equity increase (ready by 25 March)
  • We will: Send through updated estate planning referral to [Law Firm]
  • We need your approval for: Rebalancing the portfolio as discussed (formal SOA to follow)
  • You mentioned: Wanting to add your daughter as a beneficiary, we will update the nomination form

Please let us know if anything above doesn't match your understanding. We'll be in touch with the formal Statement of Advice within the next week.

Warm regards,
Steve
Lucerne Private Wealth

6. Human Verification & Approval Gates

Every AI output requires human review before any client-facing action

Multi-Gate Verification Process GATE 1: Pre-AI Sanitisation Check Automated scan confirms: - No TFN patterns remain - No known client names remain - No account numbers remain - No email/phone patterns remain If ANY found: BLOCK + alert advisor GATE 2: AI Output Quality Check Automated validation: - All tokens present in output - No hallucinated financial advice - Tone matches professional standard - Length within expected range If issues: flag for manual review GATE 3: Human Approval (MANDATORY) Advisor must personally: - Read the complete draft email - Verify accuracy of content - Edit anything that needs changing - Click explicit "Approve & Send" NOTHING sends without this click Complete Audit Trail (Every Action Logged) Input Capture What data was fetched PII Actions What was tokenised/removed AI Processing Prompt sent, response received Human Decision Approved/edited/rejected + who Logs retained for 7 years per Corporations Act record-keeping obligations (s988E)
Why Three Gates Instead of One?

Gate 1 catches PII leakage before it leaves the building. Gate 2 catches AI quality issues (hallucinations, wrong tone, missing info) before a human sees a confusing draft. Gate 3 is the final human checkpoint: the advisor who knows this client validates accuracy. This layered approach means no single point of failure can result in incorrect or privacy-violating client communications.

7. Regulatory Compliance Framework

How this solution aligns with Australian financial services regulations

Privacy Act 1988 (APP 6 & APP 11)

Compliant

APP 6 (Use & Disclosure): PII is only used for the primary purpose it was collected (servicing the client). AI processing uses tokenised data only, so no disclosure to a third party occurs.

APP 11 (Security): Encryption at rest and in transit, access controls, audit logging, and the PII never leaving the secure environment satisfies "reasonable steps" to protect personal information.

Key action: Update Lucerne's Privacy Policy to mention AI-assisted communications. Obtain opt-in client consent as best practice.

APRA CPS 234 (Information Security)

Compliant by Design

While Lucerne may not be directly APRA-regulated (they're likely ASIC-licensed), aligning with CPS 234 demonstrates best practice:

  • Information asset classification: Client PII classified as "Confidential"
  • Security controls: Commensurate with sensitivity (encryption, access control, monitoring)
  • Third-party management: Anthropic's zero-retention API means minimal third-party risk
  • Incident management: Automated PII leak detection with immediate blocking

Corporations Act 2001 (Record Keeping)

Compliant

Section 988E: Financial services licensees must keep records of financial services provided. The audit log captures every AI interaction and human approval.

Section 912A: "Efficient, honest and fair" obligation. Human review gate ensures AI outputs are verified for accuracy before client delivery.

7-year retention: All logs, drafts, approvals, and sent emails are retained for the mandated period.

ASIC RG 271 (Internal Dispute Resolution)

Consideration Required

If a client complains about an AI-generated communication, Lucerne must be able to:

  • Produce the original meeting transcript
  • Show the sanitised version sent to AI
  • Show the AI's raw output
  • Show the advisor's edits and approval
  • Show the final sent email

The audit trail satisfies this entirely. Ensure IDR procedures reference AI-assisted communications.

Client Consent Framework

Recommended: Informed Consent (Not Legally Required for This Design, But Best Practice)

Because the AI never receives PII, there is a strong argument that no additional consent is needed beyond Lucerne's existing privacy policy. However, for transparency and client trust (especially for high-net-worth individuals), we recommend:

  • Adding a clause to the client engagement letter: "We use AI-assisted tools to help prepare meeting summaries and follow-up communications. All communications are reviewed and approved by your advisor before being sent. Your personal information is not shared with AI providers."
  • Providing an opt-out mechanism for clients who prefer fully manual communications

8. Security Architecture & Data Handling

How data is protected at every stage

Anthropic / Claude API Data Handling

Zero Data Retention

When using the Claude API (not the consumer product), Anthropic does not retain or train on any data submitted. API inputs and outputs are not stored after processing. This is contractually guaranteed in Anthropic's commercial terms.

Data Processing Agreement

Anthropic offers a DPA for enterprise customers that covers GDPR, Australian Privacy Act, and other frameworks. This should be executed before go-live. Covers: data processing obligations, breach notification, sub-processor management.

SOC 2 Type II Certified

Anthropic maintains SOC 2 Type II certification, independently audited. This covers security, availability, and confidentiality controls. Audit reports can be requested for Lucerne's compliance team.

Encryption & Access Controls

LayerProtectionStandard
Data in TransitTLS 1.3 for all API calls (Microsoft Graph, Claude API)AES-256-GCM
Data at Rest (Mapping Store)AES-256 encryption for the PII token mapping databaseFIPS 140-2
Data at Rest (Audit Logs)Encrypted storage within Lucerne's Azure tenantAzure Managed Keys
API AuthenticationOAuth 2.0 with short-lived tokens, no stored passwordsRFC 6749
Claude API KeyStored in Azure Key Vault, rotated quarterlyRBAC access only
Access ControlRole-based: only authorised advisors can trigger processingAzure AD RBAC
NetworkAll processing within Lucerne's Azure VNet / tenantPrivate endpoints where possible
Data Flow Security Zones ZONE 1: LUCERNE INTERNAL (Full PII) Microsoft 365 (Emails, Meetings, CRM) PII Engine + Mapping Store Audit Logs + Rehydration Human Approval Interface BOUNDARY TLS 1.3 Tokenised data only crosses this boundary ZONE 2: ANTHROPIC (Zero PII) Claude API (Stateless Processing) Zero Data Retention SOC 2 Type II Certified DPA Available for Execution

9. Known Limitations & Mitigations

Transparency about what the system cannot do, and how we address each limitation

LimitationRisk LevelMitigation
PII detection is not 100%
Novel name formats or unusual data patterns may slip through
Medium Three-layer detection (regex + NER + dictionary) provides 99%+ coverage. Post-processing validation scan. Gate 1 automated check. Regular model retraining on missed patterns. Client name dictionary pre-loaded from CRM.
AI may hallucinate content
Claude could generate advice or details not discussed in the meeting
Medium Prompts explicitly instruct: "Only summarise what was discussed. Do not add recommendations." Gate 3 human review catches any fabricated content. Advisor knows what was actually discussed.
Teams transcript quality varies
Background noise, accents, or technical terms may be misheard
Low-Med Advisor reviews the final output against their memory of the meeting. Poor transcript quality flagged automatically based on confidence scores. Manual notes can supplement.
Financial advice generation risk
AI must not generate personal financial advice (AFSL obligation)
High System prompt explicitly prohibits advice generation. Output includes disclaimer. Only summarises what was discussed, never recommends. Gate 2 scans for advisory language patterns. Human approval mandatory.
Context-dependent PII
"My son James" may not be detected as PII without context
Low-Med NER model detects names in context. CRM lookup for known family members. Conservative approach: when in doubt, tokenise.
Data leaves Australia momentarily
Anthropic's API servers may be outside Australia
Low No PII leaves Australia (only tokenised data). Anthropic offers AWS Sydney region processing for enterprise. Zero retention means no data stored overseas. DPA covers cross-border data flow obligations.

10. Implementation Roadmap

A phased approach to reduce risk and validate at each stage

1
Foundation & Security Setup (Weeks 1-3)
2
Microsoft Integration (Weeks 3-5)
3
AI Processing & Verification (Weeks 5-7)
4
Pilot Testing (Weeks 7-10)
5
Full Rollout (Weeks 10-12)

11. ROI Perspective

The business case for AI-assisted client communications

Time Saved = Capacity Recovered

If each advisor spends 30 minutes per meeting on follow-up emails and action logging, and they have 15 meetings per week, that is 7.5 hours per advisor per week on admin. With 5 advisors, that's 37.5 hours/week. This system reduces that to ~5 minutes per meeting (review and approve), saving ~30 hours/week across the team. At an advisor billing rate of $300/hr, that's ~$9,000/week in recovered capacity, or ~$36,000/month.

30 min
Current Time Per Follow-Up
5 min
With AI (Review & Approve)
~30 hrs
Saved Per Week (5 Advisors)
~$36K
Monthly Recovered Capacity