Connect Exact Online to AI: complete practical guide

Exact Online is the accounting package for the SME. Connect its data smartly to AI, and you unlock a host of applications that cost a lot of time to do manually. This guide walks you through it step by step.

What data Exact Online has

Exact Online contains valuable data:

• Sales and purchase invoices
• General ledger and cost centers
• Debtors, creditors and outstanding items
• Inventory (in Exact Wholesale/Manufacturing)
• Project and time administration

With the right connection, this data flows to AI, which then performs analyses, forecasts and automations on it.

The Exact Online API

Exact Online offers a comprehensive REST API with OAuth 2.0 authentication. Key features:

• Available from Exact Online Standard—note: you do need an Exact developer subscription (see below)
• Rate limit: 60 calls per minute per app, 5,000 per day (per administration)
• Bulk endpoints for large datasets
• Webhooks for real-time triggers

Documentation: start.exactonline.nl/docs

Setup: step by step

Step 1: Create an Exact Online app registration

• Go to apps.exactonline.com
• Register a new app
• Note Client ID and Client Secret
• Set your redirect URL

Step 2: Implement OAuth flow

• User authorizes via Exact login page
• Your app receives an authorization code
• Exchange that for an access token and refresh token
• Refresh token remains valid as long as it's actively used

Step 3: First API call

• Retrieve divisions (administrations)
• Select the right division ID
• Query, for example, the most recent invoices

Step 4: Normalize data

• Exact uses deep nested structures
• Flatten data into workable tables
• Store for analysis in your own datawarehouse (BigQuery, Postgres, Snowflake)

Step 5: Add AI layer

Depending on the use case:

• Invoice processing: OCR + LLM for validation
• Cashflow: forecast model on historical data
• Reporting: LLM that generates reports in natural language

Top 5 AI use cases with Exact Online

1. Automatic invoice processing

Incoming invoices are read by AI, categorized, matched to purchase orders, and delivered as a draft entry. Result: 80-95% straight-through processing.

2. Cashflow forecasting

Based on outstanding invoices, debtor payment behavior and historical patterns, AI predicts cashflow for the next 30-90 days.

3. Smart reminders

AI determines the right tone, timing and channel for each debtor reminder. Result: shorter DSO (Days Sales Outstanding).

4. Anomaly detection in entries

AI flags unusual transactions, duplicates or incorrect categorizations—handy for month-end or quarter-end closing.

5. Natural language reporting

"How did our margin perform in Q3 compared to Q2?"—answered with figures, charts and context.

Security and compliance

Important points:

• Store tokens encrypted (Vault, KMS, or equivalent)
• Never give AI direct write access to production without a review step
• Document which data you extract and why (GDPR)
• Sign a DPA with your AI vendor
• Maintain audit logs for at least 30 days

Costs and timeline

A working setup:

• Exact developer subscription: around €15 per month (required for API access)
• One-time: €6,000 to €20,000 (depending on use cases)
• Monthly: €400 to €1,500 (including LLM tokens, hosting)
• First working use case: 4-6 weeks

Common mistakes

1. Not pulling history: AI needs at least 12-24 months of data for good forecasts
2. No separate sandbox: test first in an Exact test environment before writing to production
3. Underestimating rate limits: build retry logic with exponential backoff

Conclusion

Thanks to its comprehensive API, Exact Online is an excellent foundation for AI in the SMB. The combination of financial data + AI delivers immediate measurable results, especially in invoice processing and cashflow. Start with one use case, then expand from there.