Implementation Guide: Draft buyer/seller agency agreements and transaction disclosure documents

Hardware Procurement

Software Procurement

Prerequisites

• Active Microsoft 365 Business Standard (or higher) tenant with all agent/broker user accounts provisioned and licensed
• Global Admin or at least Application Admin credentials for the M365 tenant to configure Copilot add-on and SSO
• Existing CRM system (Follow Up Boss, kvCORE/BoldTrail, LionDesk, or Chime/Lofty) with active API access or Zapier integration enabled
• Valid Docusign account or willingness to procure; NAR member credentials for 20% discount
• Engagement letter with a licensed real estate attorney in the client's state(s) of operation — attorney must review and approve all document templates before production use
• List of all document types currently used by the brokerage: buyer agency agreements, seller listing agreements, property condition disclosures, lead-based paint disclosures, agency disclosure forms, and any state-specific mandated forms
• Copies of current state-approved forms from the local Realtor association or state real estate commission (e.g., via zipForms or association portal)
• Confirmation from brokerage's E&O insurance carrier that AI-assisted document drafting is covered under their current policy, or documentation of any exclusions
• Internet connectivity: minimum 25 Mbps symmetric per office location
• Domain ownership verified for SSO configuration (e.g., clientbrokerage.com)
• Identification of 3–5 pilot agents willing to test the system for 2 weeks before full rollout
• Current brokerage compliance/policy manual documenting document retention requirements and review procedures

Installation Steps

...

Step 1: Environment Assessment and Baseline Documentation

Conduct a thorough assessment of the client's current document workflow, technology stack, and compliance requirements. Interview the managing broker and 2–3 agents to understand pain points, typical transaction volume, state-specific form requirements, and existing software. Document the current state in a findings report that will guide all subsequent configuration decisions.

Step 2: Microsoft 365 Tenant Configuration and Copilot Deployment

Verify the client's M365 tenant is on Business Standard or higher. Assign Microsoft 365 Copilot add-on licenses to all agent/broker users. Configure Conditional Access policies to require MFA for all users. Create a SharePoint site named 'Transaction Documents' with document libraries organized by year and transaction type. Configure retention policies matching state requirements (minimum 3 years, up to 7 years depending on state).

# Connect to Microsoft Graph PowerShell
Connect-MgGraph -Scopes 'User.ReadWrite.All','Organization.Read.All','Directory.ReadWrite.All'
# Verify tenant subscription
Get-MgSubscribedSku | Select-Object SkuPartNumber, ConsumedUnits, PrepaidUnits
# Assign Copilot licenses (replace with actual User IDs)
$copilotSkuId = (Get-MgSubscribedSku | Where-Object {$_.SkuPartNumber -eq 'Microsoft_365_Copilot'}).SkuId
$users = Get-MgUser -Filter "department eq 'Sales'" -All
foreach ($user in $users) {
    Set-MgUserLicense -UserId $user.Id -AddLicenses @(@{SkuId=$copilotSkuId}) -RemoveLicenses @()
}
# Create SharePoint site for transaction documents
New-MgSite -DisplayName 'Transaction Documents' -Description 'AI-generated and executed real estate documents' -WebUrl 'https://clientbrokerage.sharepoint.com/sites/TransactionDocs'
# Configure retention policy via Compliance Center PowerShell
Connect-IPPSSession
New-RetentionCompliancePolicy -Name 'Real Estate Document Retention' -SharePointLocation 'https://clientbrokerage.sharepoint.com/sites/TransactionDocs' -Enabled $true
New-RetentionComplianceRule -Policy 'Real Estate Document Retention' -RetentionDuration 2555 -RetentionComplianceAction Keep -Name '7-Year Retention Rule'

Step 3: Gavel Platform Setup and Organization Configuration

Create the brokerage's Gavel organization account. Configure SSO via Microsoft Entra ID (Azure AD) so agents authenticate with their M365 credentials. Purchase 2 Builder licenses (for the managing broker or office admin and the MSP administrator) and configure viewer access for agents. Set up the document output destinations (SharePoint and local download).

Step 4: Document Template Development — Buyer Agency Agreement

Build the first and most critical template: the Buyer Agency Agreement compliant with the August 2024 NAR settlement requirements. In Gavel's workflow builder, create a guided questionnaire that collects all required information, then auto-populates a Word template with conditional logic for state-specific variations. The template MUST include: (1) specific and conspicuous disclosure of agent compensation amount or rate, (2) how compensation will be determined if not a fixed amount, (3) all contractual obligations of the buyer, (4) duties of confidentiality, (5) Equal Housing Opportunity statement, (6) agreement duration and termination provisions, and (7) state-specific mandatory disclosures.

# Gavel Workflow Questionnaire Fields (configure in Gavel's visual builder):
# --- Buyer Information ---
# buyer_full_name (text, required)
# buyer_email (email, required)
# buyer_phone (phone, required)
# buyer_mailing_address (address, required)
# co_buyer_full_name (text, optional)
# co_buyer_email (email, optional)
# --- Agent/Brokerage Information ---
# agent_full_name (text, required)
# agent_license_number (text, required)
# brokerage_name (text, required, pre-filled)
# brokerage_license_number (text, required, pre-filled)
# brokerage_address (address, required, pre-filled)
# supervising_broker_name (text, required, pre-filled)
# --- Compensation Disclosure (NAR Settlement Required) ---
# compensation_type (dropdown: 'Fixed Dollar Amount' | 'Percentage of Purchase Price' | 'Hourly Rate' | 'Other')
# compensation_amount (text, required — e.g., '$5,000' or '2.5%')
# compensation_determination_method (textarea, conditional — shown if compensation_type = 'Other')
# compensation_paid_by (dropdown: 'Buyer' | 'Seller (if offered)' | 'Split')
# --- Agreement Terms ---
# agreement_start_date (date, required)
# agreement_end_date (date, required)
# geographic_area (text, required — e.g., 'Denver Metro Area')
# property_type (multi-select: 'Single Family' | 'Condo/Townhome' | 'Multi-Family' | 'Land' | 'Commercial')
# price_range_low (currency, optional)
# price_range_high (currency, optional)
# exclusive_or_nonexclusive (dropdown: 'Exclusive' | 'Non-Exclusive')
# --- State Selection (drives conditional clauses) ---
# operating_state (dropdown: list of all 50 states + DC)
# --- Termination ---
# termination_notice_days (number, default: 30)
# early_termination_fee (currency, optional)
# --- Additional Terms ---
# additional_terms (textarea, optional)

Step 5: Document Template Development — Seller Listing Agreement

Build the Seller Listing Agreement template in Gavel. This template collects property details, listing terms, commission structure, marketing authorizations, and state-specific disclosures. Include conditional sections for: listing price strategy (fixed vs. auction), dual agency disclosure (where permitted by state), lockbox authorization, photography/virtual tour consent, and MLS data entry authorization.

• Seller Information: seller_full_name (text, required), seller_email (email, required), seller_phone (phone, required), co_seller_full_name (text, optional), seller_current_address (address, required)
• Property Information: property_address (address, required), property_type (dropdown: 'Single Family' | 'Condo' | 'Townhome' | 'Multi-Family' | 'Land' | 'Commercial'), legal_description (textarea, required), parcel_number (text, required)
• Listing Terms: listing_price (currency, required), listing_start_date (date, required), listing_expiration_date (date, required), listing_type (dropdown: 'Exclusive Right to Sell' | 'Exclusive Agency' | 'Open Listing')
• Commission Structure: listing_commission_rate (text, required — e.g., '5%' or '$15,000'), buyer_agent_commission_offered (text, required — post-NAR settlement this is optional but must be documented), commission_paid_at (dropdown: 'Closing' | 'Other')
• Marketing Authorizations: authorize_mls_listing (boolean, default: true), authorize_photography (boolean, default: true), authorize_virtual_tour (boolean, default: true), authorize_open_houses (boolean, default: true), authorize_social_media (boolean, default: true), lockbox_authorized (boolean, default: true)
• State-Specific: operating_state (dropdown), dual_agency_permitted (boolean, conditional on state), dual_agency_consent (boolean, conditional)
• Additional Terms: personal_property_included (textarea, optional), fixtures_excluded (textarea, optional), additional_terms (textarea, optional)

Step 6: Document Template Development — Transaction Disclosure Documents

Build templates for the most commonly required transaction disclosure documents. At minimum, create templates for: (1) Residential Property Condition Disclosure (state-specific), (2) Lead-Based Paint Disclosure (federal — required for pre-1978 properties), (3) Agency Relationship Disclosure, and (4) Wire Fraud Warning. Each disclosure template should use Gavel's conditional logic to show/hide sections based on the operating state and property characteristics.

# Lead-Based Paint Disclosure — Federal Template (42 U.S.C. 4852d)
# This is a standardized federal form — template must match the prescribed format exactly.
# Gavel fields:
# property_address (text, required)
# property_built_before_1978 (boolean, required)
#   IF true → show full lead-based paint disclosure form
#   IF false → show exemption acknowledgment
# seller_knowledge_of_lead_paint (dropdown: 'Known lead-based paint present' | 'No knowledge of lead-based paint' | 'Records/reports available')
# seller_explanation (textarea, conditional on 'Known lead-based paint present')
# available_records_description (textarea, conditional on 'Records/reports available')
# buyer_received_epa_pamphlet (boolean, required — must be true to proceed)
# buyer_10_day_inspection_opportunity (dropdown: 'Buyer elects 10-day inspection period' | 'Buyer waives 10-day inspection period')
# buyer_signature_date (date)
# seller_signature_date (date)
# agent_signature_date (date)

# Property Condition Disclosure — Configurable by State
# operating_state drives which sections appear:
# IF state = 'California' → use TDS (Transfer Disclosure Statement) format with specific CA Civil Code sections
# IF state = 'Texas' → use TREC Seller's Disclosure of Property Condition
# IF state = 'New York' → offer $500 credit option (Property Condition Disclosure Act)
# Each state variant maintained as a separate Gavel sub-workflow linked via the state selector

Step 7: Configure CRM-to-Gavel Integration via Zapier

Set up Zapier Professional to connect the client's CRM (Follow Up Boss used in this example) to Gavel, enabling automatic pre-population of document fields when a new transaction is initiated. Create a multi-step Zap: Trigger → CRM new deal/transaction created → Lookup contact details → Send data to Gavel via webhook → Gavel creates pre-populated questionnaire link → Send link to assigned agent via email/Slack.

# Zapier Configuration — Zap 1: New Transaction → Gavel Pre-Population
# Step 1 (Trigger): Follow Up Boss — New Deal Created
#   - Connect Follow Up Boss account via API key
#   - Trigger event: New Deal
#   - Filter: Deal Stage = 'Active' or 'Under Contract'

# Step 2 (Action): Follow Up Boss — Find Person
#   - Lookup the contact associated with the deal
#   - Map: contact_id from Step 1

# Step 3 (Action): Webhooks by Zapier — POST to Gavel API
#   - URL: https://api.gavel.io/v1/workflows/{workflow_id}/submissions
#   - Method: POST
#   - Headers: Authorization: Bearer {GAVEL_API_KEY}
#   - Body (JSON):
#     {
#       "buyer_full_name": "{{Step2_FirstName}} {{Step2_LastName}}",
#       "buyer_email": "{{Step2_Email}}",
#       "buyer_phone": "{{Step2_Phone}}",
#       "agent_full_name": "{{Step1_AssignedAgent}}",
#       "property_address": "{{Step1_DealAddress}}",
#       "operating_state": "{{Step1_DealState}}"
#     }

# Step 4 (Action): Gmail/Outlook — Send Email to Agent
#   - To: {{Step1_AssignedAgentEmail}}
#   - Subject: 'New Transaction Document Draft Ready — {{Step1_DealAddress}}'
#   - Body: 'A new buyer agency agreement draft has been pre-populated for {{Step2_FirstName}} {{Step2_LastName}} at {{Step1_DealAddress}}. Review and complete the questionnaire here: {{Step3_SubmissionURL}}'

# Zapier Configuration — Zap 2: Gavel Document Generated → Docusign Envelope
# Step 1 (Trigger): Webhooks by Zapier — Catch Hook
#   - Gavel sends webhook on document generation completion
# Step 2 (Action): Docusign — Create Envelope from Template
#   - Map document file from Step 1 webhook payload
#   - Set signers from Gavel output fields
#   - Send for signature

Step 8: Configure Docusign Rooms for Real Estate

Set up Docusign Rooms for Real Estate to serve as the centralized transaction workspace. Create room templates for each transaction type (buyer-side, seller-side, dual) that automatically organize generated documents into the correct compliance folders. Configure the document library with all approved templates. Set up role-based access so agents see their transactions, managing broker sees all transactions, and the MSP admin has system configuration access.

Step 9: OpenAI API Configuration for Supplementary AI Generation

Set up an OpenAI API account for custom document generation needs not covered by Gavel templates — such as generating custom addendum language, plain-English summaries of disclosure documents for clients, or handling unusual transaction types. Configure API access with usage limits, and deploy a simple internal web tool or Microsoft Power Automate flow that agents can use to request custom document content with appropriate guardrails.

# 1. Create OpenAI API account at https://platform.openai.com
# 2. Generate API key: Settings → API Keys → Create new secret key
#    Name: 'ClientBrokerage-DocGen-Production'
#    Permissions: 'Chat Completions only'
# 3. Set usage limits: Settings → Billing → Usage Limits
#    Monthly budget: $100 (sufficient for ~2,000 document generations)
#    Alert threshold: $75
# 4. Store API key in Azure Key Vault or client's password manager (never in plain text)

# Power Automate Flow Configuration:
# Trigger: Manual trigger (agent clicks 'Generate Custom Addendum' in Teams)
# Input fields: transaction_address, addendum_type, specific_terms, state
# Action: HTTP POST to https://api.openai.com/v1/chat/completions
# Headers: Authorization: Bearer {{OpenAI_API_Key}}, Content-Type: application/json
# Body: See custom_ai_components section for full prompt template
# Action: Send response to requesting agent via Teams adaptive card
# Action: Log generation to SharePoint list for audit trail

Step 10: Implement Mandatory Human Review Workflow

Configure a mandatory human-in-the-loop review checkpoint that prevents any AI-generated document from being sent for signature without explicit agent and/or managing broker review. This is the most critical compliance control in the entire system — it mitigates UPL risk, ensures document accuracy, and provides an auditable approval trail.

• SharePoint List: 'Document Review Log' — Columns: DocumentID (auto-number), TransactionAddress (text), DocumentType (choice: Buyer Agency | Seller Listing | Disclosure | Addendum | Other), GeneratedBy (text: 'Gavel Template' | 'OpenAI API' | 'Copilot Assist'), GeneratedDate (datetime), ReviewedByAgent (person), AgentReviewDate (datetime), AgentApproved (yes/no), ReviewedByBroker (person, optional — required for certain document types), BrokerReviewDate (datetime, optional), BrokerApproved (yes/no, optional), SentForSignature (yes/no), SignatureEnvelopeID (text — Docusign envelope ID), Notes (multi-line text)
• Power Automate Flow: 'Document Review Gate' — Trigger: When a new item is created in 'Document Review Log'
• Condition: If DocumentType = 'Buyer Agency Agreement' OR 'Seller Listing Agreement' → Send approval request to Managing Broker → Wait for approval (timeout: 48 hours) → If approved: Update BrokerApproved = Yes, proceed to Docusign → If rejected: Notify agent with rejection reason, log rejection
• Condition: If DocumentType = 'Disclosure' or 'Addendum' → Agent self-review is sufficient (no broker approval required) → Agent marks AgentApproved = Yes → proceed to Docusign

Step 11: Security Hardening and Access Controls

Apply security controls appropriate for a system handling PII and legal documents. Configure Entra ID Conditional Access policies, enable DLP (Data Loss Prevention) policies for the Transaction Documents SharePoint site, and set up audit logging.

# Entra ID Conditional Access Policy
Connect-MgGraph -Scopes 'Policy.ReadWrite.ConditionalAccess'

• Conditional Access Policy: Require MFA for all Gavel, DocuSign, and SharePoint Transaction Documents access
• Conditional Access Policy: Block access from non-compliant devices
• Conditional Access Policy: Require managed device for SharePoint Transaction Documents site

# Enable unified audit logging (if not already enabled):
Set-AdminAuditLogConfig -UnifiedAuditLogIngestionEnabled $true
# Verify:
Get-AdminAuditLogConfig | Select-Object UnifiedAuditLogIngestionEnabled

Step 12: Pilot Deployment with Test Agents

Deploy the system to 3–5 pilot agents for a 2-week monitored trial period. Each pilot agent processes at least 2–3 real transactions through the system while continuing to use their existing document workflow in parallel (belt-and-suspenders approach). The MSP monitors system performance, collects feedback, and iterates on templates and workflows.

Step 13: Full Rollout and Agent Training

After successful pilot completion, roll out to all agents. Conduct 2–3 training sessions (in-person or video) covering: how to initiate a new document from the CRM, how to complete the Gavel questionnaire, how to review and edit AI-generated drafts in Word using Copilot, how to submit documents for broker review, and how to send documents for eSignature via Docusign. Provide written quick-reference guides and short video recordings of each workflow.

Custom AI Components

Buyer Agency Agreement Generator Prompt

Type: prompt System prompt for generating buyer agency agreement content via the OpenAI API when custom language is needed beyond Gavel's template capabilities. This prompt ensures NAR settlement compliance, state-specific requirements, and appropriate disclaimers. Used primarily for generating custom additional terms, special conditions, or non-standard agreement structures.

Implementation:

Model: gpt-4.1
Temperature: 0.2 (low creativity for legal precision)
Max tokens: 4096
Top_p: 0.95

Transaction Disclosure Summary Generator

Type: prompt Generates plain-English summaries of complex disclosure documents for buyer/seller communication. Real estate clients often struggle to understand legal disclosure language — this component creates a companion summary that agents can share alongside the formal disclosure document to improve client comprehension and reduce follow-up questions.

Implementation

Model: gpt-4.1
Temperature: 0.3
Max tokens: 2048
Top_p: 0.95

Custom Addendum Drafting Assistant

Type: prompt Generates custom addendum and amendment language for real estate transactions when standard templates don't cover the specific situation. Common use cases: inspection contingency modifications, financing contingency extensions, repair request addenda, closing date changes, and personal property inclusion/exclusion addenda.

Implementation:

Model: gpt-4.1
Temperature: 0.15 (very low — maximum precision for legal modifications)
Max tokens: 3072
Top_p: 0.9

CRM-to-Gavel Auto-Population Workflow

Type: workflow Zapier-based automation workflow that connects Follow Up Boss CRM to Gavel document automation platform. When an agent creates a new deal in the CRM, this workflow automatically pre-populates a Gavel questionnaire with buyer/seller data and sends the agent a link to review, complete, and generate the appropriate documents.

Implementation:

WORKFLOW: CRM-to-Gavel Document Pre-Population
PLATFORM: Zapier Professional
ESTIMATED TASKS PER EXECUTION: 4

STEP 1 — TRIGGER
  App: Follow Up Boss
  Event: New Deal Created
  Configuration:
    - Account: [Client's FUB account]
    - Trigger on: All new deals
  Output fields used: deal_id, deal_name, deal_stage, deal_address, deal_city, deal_state, deal_zip, assigned_agent_name, assigned_agent_email, contact_id

STEP 2 — LOOKUP
  App: Follow Up Boss
  Event: Find Person
  Configuration:
    - Search by: contact_id from Step 1
  Output fields used: first_name, last_name, email, phone, mailing_address

STEP 3 — FILTER
  Condition: Only continue if deal_stage = 'New' OR deal_stage = 'Active Buyer' OR deal_stage = 'Listing Appointment'
  Purpose: Prevent duplicate document generation for deals that have already progressed past initial document stage

STEP 4 — ROUTER
  Path A: If deal_stage contains 'Buyer' → Use Buyer Agency Agreement workflow_id
  Path B: If deal_stage contains 'Listing' or 'Seller' → Use Seller Listing Agreement workflow_id
  Path C: Default → Use General Document Selection workflow_id

STEP 5 — ACTION (for each path)
  App: Webhooks by Zapier
  Event: POST
  Configuration:
    URL: https://api.gavel.io/v1/workflows/{{workflow_id}}/submissions
    Method: POST
    Headers:
      Authorization: Bearer {{GAVEL_API_KEY}}
      Content-Type: application/json
    Body:
    {
      "prefill": {
        "buyer_full_name": "{{Step2.first_name}} {{Step2.last_name}}",
        "buyer_email": "{{Step2.email}}",
        "buyer_phone": "{{Step2.phone}}",
        "property_address": "{{Step1.deal_address}}, {{Step1.deal_city}}, {{Step1.deal_state}} {{Step1.deal_zip}}",
        "operating_state": "{{Step1.deal_state}}",
        "agent_full_name": "{{Step1.assigned_agent_name}}"
      },
      "notify": false
    }
  Output: submission_url

STEP 6 — ACTION
  App: Microsoft Outlook (or Gmail)
  Event: Send Email
  Configuration:
    To: {{Step1.assigned_agent_email}}
    Subject: "📄 New Document Draft Ready: {{Step1.deal_address}}"
    Body: "
      Hi {{Step1.assigned_agent_name}},

      A new {{Path A: 'Buyer Agency Agreement' | Path B: 'Seller Listing Agreement'}} has been pre-populated for:

      Client: {{Step2.first_name}} {{Step2.last_name}}
      Property: {{Step1.deal_address}}

      Click here to review and complete the document: {{Step5.submission_url}}

      REMINDER: Review all pre-filled information for accuracy before generating the final document. All generated documents require your review before sending for signature.

      Questions? Contact [MSP Support Email/Phone]
    "

STEP 7 — ACTION
  App: Microsoft SharePoint
  Event: Create List Item
  Configuration:
    Site: Transaction Documents
    List: Document Generation Log
    Fields:
      TransactionAddress: {{Step1.deal_address}}
      DocumentType: {{Router path result}}
      GeneratedBy: 'Gavel Template (auto-triggered)'
      GeneratedDate: {{current_timestamp}}
      ReviewedByAgent: (blank — pending)
      AgentApproved: No

MONITORING:
- Check Zapier Task History weekly for failed tasks
- Set up Zapier email alerts for any task failures
- Review Document Generation Log monthly for audit compliance

Document Compliance Checker

Type: prompt A quality assurance prompt that reviews AI-generated documents against a compliance checklist specific to the document type and state. Used by the managing broker or compliance officer as a final check before documents are sent for signature.

Implementation:

Model: gpt-4.1 (1M context window supports full document review)
Temperature: 0.1 (maximum precision for compliance checking)
Max tokens: 4096
Top_p: 0.9

Fair Housing Language Auditor

Type: prompt Specialized prompt that audits document templates and generated documents for potential Fair Housing Act violations. Checks for discriminatory language, steering implications, or exclusionary terms related to the seven protected classes (race, color, national origin, religion, sex, disability, familial status).

Implementation:

API CALL CONFIGURATION:
Model: gpt-4.1
Temperature: 0.1
Max tokens: 2048

Testing & Validation

• TEST 1 — Gavel Template Accuracy: Generate a Buyer Agency Agreement for a test transaction in the client's primary state. Verify all questionnaire fields correctly populate in the output document. Check that the compensation disclosure section is bold and bordered. Compare output against the state association's standard form to ensure no required elements are missing. PASS CRITERIA: 100% of required fields populated correctly; all NAR settlement disclosures present.
• TEST 2 — State-Specific Conditional Logic: Run the Buyer Agency Agreement template for 3 different states (client's primary state + 2 others). Verify that state-specific clauses appear/disappear correctly based on the state selection. PASS CRITERIA: Each state output contains only applicable state-specific language; no cross-state contamination.
• TEST 3 — CRM-to-Gavel Integration: Create a test deal in Follow Up Boss with known buyer data. Verify the Zapier workflow triggers, pre-populates the Gavel questionnaire with correct buyer name, email, phone, property address, and state, and sends the notification email to the assigned agent within 5 minutes. PASS CRITERIA: All 6 data fields correctly mapped; email received within 5 minutes; no Zapier task errors.
• TEST 4 — Docusign Envelope Creation: Generate a complete Buyer Agency Agreement through Gavel and verify it creates a Docusign envelope with correct signer information, signing order, and signature/initial placement. Send a test envelope and complete the signing process. PASS CRITERIA: Envelope created with correct signers; all signature and initial fields properly placed; completed envelope stored in Docusign Rooms.
• TEST 5 — Human Review Gate: Generate a test document and verify that the Document Review Log entry is created in SharePoint, the approval request is sent to the managing broker via Power Automate, and the document CANNOT proceed to Docusign until the broker approves. Test rejection flow to confirm the agent receives rejection notification with reason. PASS CRITERIA: Document blocked from signature without approval; approval/rejection notifications sent correctly; audit log accurately reflects all actions.
• TEST 6 — OpenAI API Custom Addendum: Use the Custom Addendum Drafting Assistant prompt to generate an inspection contingency extension addendum. Verify the output includes all required elements: AI draft header/disclaimer, original agreement reference, property address, party names, specific modification language, integration clause, and signature blocks. PASS CRITERIA: All required elements present; language is clear and unambiguous; AI disclaimer prominently displayed.
• TEST 7 — Compliance Checker Accuracy: Run the Document Compliance Checker prompt against a known-deficient test document (deliberately missing the compensation disclosure and Equal Housing statement). Verify the checker identifies both missing items as FAILED. Then run against a compliant document and verify it returns PASS. PASS CRITERIA: 100% detection rate on deliberately missing required elements; no false failures on compliant documents.
• TEST 8 — Fair Housing Audit: Run the Fair Housing Language Auditor against a test document containing deliberately inserted problematic language (e.g., 'ideal for young professionals' — familial status implication, 'close to [specific church]' — religious preference). Verify detection. PASS CRITERIA: All deliberately inserted issues flagged as moderate or high risk.
• TEST 9 — Document Retention and Backup: Generate a document, complete the signature process, and verify the executed document is stored in (a) Docusign Rooms, (b) SharePoint Transaction Documents library, and (c) Dropsuite backup captures the SharePoint file within 24 hours. PASS CRITERIA: Document present in all three locations; SharePoint retention policy correctly applied.
• TEST 10 — End-to-End Transaction Workflow: Process a complete mock transaction from CRM deal creation through document generation, review, signature, and archival. Time the entire workflow. PASS CRITERIA: Total time from deal creation to signature-ready document under 15 minutes (vs. typical 45–90 minutes manual); zero manual data re-entry required; complete audit trail in SharePoint.
• TEST 11 — Load Test: Have 3 agents simultaneously generate documents through the system to verify no conflicts, queue delays, or data cross-contamination between concurrent submissions. PASS CRITERIA: All 3 documents generated correctly with no data from one transaction appearing in another; all completed within 5 minutes.
• TEST 12 — Attorney Review Sign-Off: Present 5 sample documents (1 of each type) generated by the system to the reviewing attorney. Obtain written confirmation that templates meet state legal requirements and are suitable for use with human review. PASS CRITERIA: Attorney provides written approval of all 5 templates with any required modifications documented and implemented.

Client Handoff

Maintenance

ONGOING MAINTENANCE RESPONSIBILITIES:

1. Monthly Tasks (4–8 hours/month)

• Review Zapier task logs for integration errors; resolve any failed tasks within 24 hours
• Review Document Generation Log in SharePoint for anomalies (documents generated but never reviewed, unusually high generation volumes from single agents)
• Check OpenAI API usage dashboard against budget; adjust limits if needed
• Verify Dropsuite backup completion for Transaction Documents SharePoint site
• Apply any pending M365, Gavel, or Docusign updates/patches
• Respond to agent support tickets (average 3–5 per month during first 6 months, declining to 1–2 per month thereafter)
• Monthly 30-minute check-in call with managing broker to review system health and upcoming needs

2. Quarterly Tasks (8–12 hours/quarter)

• Run the Document Compliance Checker against 5–10 randomly selected documents generated in the prior quarter
• Run the Fair Housing Language Auditor against all active templates
• Review NAR, state real estate commission, and CFPB updates for any regulatory changes affecting document requirements
• Update Gavel templates to reflect any state form changes (state associations typically update forms annually or semi-annually)
• Coordinate with the reviewing attorney for any template modifications requiring legal sign-off
• Review and rotate OpenAI API keys
• Conduct a brief (15-minute) user satisfaction survey with agents
• Produce a quarterly report for the managing broker: documents generated, time savings, compliance audit results, system health

3. Annual Tasks

• Full review and update of all document templates against current state forms and regulations
• Review software licensing: optimize seat counts, assess plan tier adequacy (Zapier task volume, Gavel builder licenses, etc.)
• Renew attorney engagement letter for template review
• Verify E&O insurance coverage still includes AI-assisted document workflows
• Assess new AI model versions (e.g., GPT-5, Claude next-gen) for potential quality or cost improvements
• Review and update Conditional Access and DLP policies in M365
• Conduct a full disaster recovery test: restore documents from Dropsuite backup to verify integrity

4. Trigger-Based Updates (as needed)

• NAR policy changes → immediate template review and update (target: within 2 weeks of effective date)
• State form updates → template update within 30 days of new form availability
• AI model deprecation notices → migrate to replacement model within vendor's deprecation timeline
• Security incidents → immediate response per MSP incident response SLA; notify brokerage managing broker within 1 hour
• Client expansion to new state → new state template development (budget 15–25 hours + attorney review per state)

5. SLA Considerations

• System availability: Target 99.5% uptime (dependent on SaaS vendor SLAs)
• Support response: 4-hour response for P1 (system down), 8-hour for P2 (degraded functionality), next business day for P3 (questions/enhancements)
• Template updates for regulatory changes: Within 2 weeks of notification
• Integration break-fix: Within 24 hours

6. Escalation Paths

• Level 1: MSP helpdesk (agent questions, minor issues)
• Level 2: MSP senior technician (integration failures, template errors)
• Level 3: MSP solutions architect + reviewing attorney (compliance issues, regulatory changes)
• Vendor escalation: Gavel support, Docusign support, OpenAI developer support, Microsoft support

Alternatives

...

Spellbook + Microsoft Word Native Approach

Instead of Gavel as the primary document automation platform, use Spellbook (AI-powered contract drafting add-in for Microsoft Word) combined with Word templates and Microsoft 365 Copilot. Agents work entirely within Word — Spellbook provides AI clause suggestions, contract review, and redlining directly in the document editor. Templates are stored as Word documents in SharePoint with content controls for variable fields.

Dotloop or SkySlope All-in-One Platform

Instead of building a multi-vendor integration stack, use Dotloop Premium ($31.99/agent/month) or SkySlope Pro ($40/agent/month) as the primary platform for document management, forms, eSignature, and compliance — then add Microsoft 365 Copilot for AI-assisted drafting within the platform's existing document workflows. This minimizes the number of integrations and vendors.

Custom-Built Solution with OpenAI API + Docassemble

Build a fully custom document generation pipeline using OpenAI GPT-4.1 API for content generation, Docassemble (open-source) for guided interview-based document assembly, and custom web interface hosted on Azure or AWS. The MSP owns and controls the entire stack, providing maximum customization and the highest margin opportunity.

Microsoft 365 Copilot Only (Minimal Approach)

The simplest possible deployment: add Microsoft 365 Copilot licenses to existing M365 accounts and train agents to use Copilot in Word for document drafting assistance. No additional SaaS platforms, no integrations, no custom development. Agents use Copilot to help draft documents from existing Word templates stored in SharePoint.

On-Premise LLM Deployment (Maximum Privacy)

For brokerages with strict data sovereignty requirements or operating in states with stringent data privacy laws, deploy an on-premise LLM (Llama 3.3 70B or Mistral Large via Ollama) on a local server. All document generation happens on-site — no client PII is sent to cloud APIs. Combined with Docassemble for document assembly and local document storage.