Operations
January 2025
·
7 min read

How to Prepare Your ERP for AI Agent Implementation

The Preparation Question

AI agents require ERP access to read data and write outcomes. Understanding preparation requirements prevents implementation delays and ensures smooth integration.

Most modern ERPs are ready. Some require minor preparation. A few need significant work.

Technical Requirements

API Access

What's needed:

  • REST API or similar modern interface
  • Authentication capability (OAuth preferred)
  • Read access to exception data
  • Write access for outcomes and notes

Standard ERPs ready:

  • Acumatica: Comprehensive REST API
  • NetSuite: SuiteTalk REST/SOAP APIs
  • Dynamics 365: Web API (OData)
  • SAP Business One: Service Layer
  • Sage Intacct: Web Services API

Custom or legacy ERPs: May require API development or database direct access (less preferred)

Data Access Permissions

Read permissions required:

  • Customer master data
  • Invoice/transaction data
  • Payment history
  • Account status and balances
  • Contact information
  • Notes and activity history

Write permissions required:

  • Activity logging (calls, emails)
  • Notes and comments
  • Status updates
  • Task creation
  • Payment commitments

Data Quality Requirements

Critical Data Elements

Customer contact information:

  • Phone numbers (valid format)
  • Email addresses (clean, deliverable)
  • Contact names (decision makers)
  • Preferred communication method

Quality check: 80%+ of accounts have valid phone/email

Transaction data:

  • Invoice numbers
  • Amounts and dates
  • Due dates
  • Current balance
  • Payment terms

Quality check: Complete and accurate for all open items

Common Data Issues

Issue 1: Missing Contact Information

Symptom: 30-40% of accounts lack phone numbers

Impact: AI cannot contact these accounts

Solution: Data cleanup project before implementation. Target: 80%+ coverage.

Timeline: 2-4 weeks

Cost: $0 (internal) to $3,000 (consultant assistance)

Issue 2: Duplicate Records

Symptom: Same customer appears multiple times

Impact: AI may contact wrong account or miss communications

Solution: Data deduplication exercise

Timeline: 1-3 weeks

Cost: $0-$5,000 depending on volume

Issue 3: Stale or Incorrect Data

Symptom: Phone numbers disconnected, emails bouncing

Impact: Failed contact attempts, poor success rates

Solution: Ongoing data validation process

Timeline: Continuous improvement

Cost: Included in oversight time

Process Documentation

Current State Documentation

Required documentation:

  • How exceptions are currently identified
  • Who handles each exception type
  • Decision criteria for actions
  • Escalation rules
  • Standard communication templates

Purpose: Translates human knowledge into AI rules

Timeline: 2-3 workshops, 6-8 hours total

Cost: Included in implementation

Decision Rules

Must document:

  • When to call vs email
  • Payment plan criteria and limits
  • Dispute handling approach
  • VIP account identification
  • Escalation triggers

Format: Decision trees or if-then rules

Example:

IF account balance > $10,000 AND overdue > 45 days

THEN call before email

IF customer is VIP

THEN escalate to controller (no AI contact)

IF customer reports dispute

THEN escalate to AR manager with dispute details

Technical Assessment

Pre-Implementation Checklist

1. API Availability

  • [ ] REST API exists and documented
  • [ ] OAuth or API key authentication available
  • [ ] Test environment accessible
  • [ ] Rate limits understood (calls per hour)

2. Data Quality

  • [ ] 80%+ accounts have valid contact info
  • [ ] Customer data reasonably clean
  • [ ] Transaction data accurate and complete
  • [ ] No major duplicate record issues

3. Security and Permissions

  • [ ] Service account can be created
  • [ ] Read/write permissions can be granted
  • [ ] Security team approves external integration
  • [ ] Data privacy requirements understood

4. IT Resource Availability

  • [ ] IT liaison identified (5-10 hours during implementation)
  • [ ] Access to test environment provided
  • [ ] Production credentials can be obtained
  • [ ] IT team available for troubleshooting

Standard ERP Configurations

Acumatica

Typical readiness: 90%+ implementations require no additional preparation

API capabilities:

  • Comprehensive REST API (Contract-Based)
  • OAuth 2.0 authentication
  • Custom fields easily added
  • Webhooks for real-time updates

Common additions:

  • Custom fields for AI interaction tracking
  • Workflow for AI-created tasks
  • Custom inquiry for AI dashboard

Timeline: 1-2 days configuration

Cost: Included in implementation

NetSuite

Typical readiness: 85% implementations ready, 15% need minor work

API capabilities:

  • SuiteTalk REST and SOAP APIs
  • Token-based authentication
  • Scripting for custom logic
  • Saved searches for data access

Common needs:

  • Custom fields for AI outcomes
  • Saved searches for exception identification
  • Script deployment for specific workflows

Timeline: 2-4 days configuration

Cost: Included in implementation

Dynamics 365

Typical readiness: 85% implementations ready

API capabilities:

  • Web API (OData)
  • Azure AD authentication
  • Custom entities supported
  • Business rules platform

Common additions:

  • Custom fields/entities
  • Security roles for service account
  • Business rules for AI workflows

Timeline: 2-3 days configuration

Cost: Included in implementation

Custom or Legacy ERP Preparation

Assessment Questions

1. Does an API exist?

  • Yes, documented: Standard implementation ($0 additional)
  • Yes, undocumented: Testing required ($2K-$5K)
  • No: API development needed ($15K-$40K) or database access alternative

2. Can API access customer and transaction data?

  • All data: Ready
  • Partial data: Custom development ($5K-$15K)
  • Limited access: May require database direct access

3. Can API write outcomes back?

  • Yes: Ready
  • No: Custom development or manual outcome entry

Database Direct Access Alternative

When API insufficient: Direct database access possible but not preferred

Requirements:

  • Read-only replica or reporting database
  • ODBC/JDBC connectivity
  • Documented table schema
  • Write-back via flat file import or similar

Considerations:

  • Less real-time than API
  • More brittle (schema changes break integration)
  • Security concerns (broader access)
  • Additional development cost ($10K-$25K)

Timeline and Cost

Standard ERP (Acumatica, NetSuite, Dynamics)

Preparation timeline: 1-2 weeks Preparation cost: $0-$5,000 (minor data cleanup if needed) Implementation timeline: 8-12 weeks total Total cost: $35,000-$45,000 including preparation

Custom ERP with Good API

Preparation timeline: 2-4 weeks Preparation cost: $5,000-$15,000 (API testing, custom development) Implementation timeline: 10-14 weeks total Total cost: $40,000-$60,000

Custom ERP with Limited/No API

Preparation timeline: 6-12 weeks Preparation cost: $15,000-$40,000 (API development or database access) Implementation timeline: 14-20 weeks total Total cost: $50,000-$85,000

Alternative: Consider ERP upgrade before AI implementation

The Reality

Most modern ERPs (Acumatica, NetSuite, Dynamics) require minimal preparation. API access exists, data quality is acceptable, minor configuration only.

Custom or legacy ERPs may need 2-12 weeks preparation including API development, data cleanup, or alternative integration approaches.

Pre-implementation assessment identifies requirements early. Budget and timeline adjust based on findings.

Standard ERPs: 1-2 weeks, $0-$5K preparation. Custom ERPs: 2-12 weeks, $5K-$40K preparation.

ABOUT THE AUTHOR

This content is published by ERP AI Agent.

ON THIS PAGE

Related Articles

See All →