Financial data migrations are complex, often taking months or years to complete. Doyen is dramatically accelerating migrations using AI while keeping them accurate, auditable, and trustworthy. With Doyen, AI is increasingly generating the migration logic by working in conjunction with finance and business professionals. But transparency becomes even more critical the more work the AI does. New paradigms for how humans interact with AI are emerging. This capability is an additional feature which highlights the shift towards humans and AI working collaboratively to solve difficult technical challenges.
Why This Matters
Whenever AI is used, it’s critical to have tools that ensure what the AI is doing is correct. The underlying models (LLMs in particular) are probabilistic in nature. To keep migrations reliable, we’ve already built multiple safeguards including reconciliation dashboards, validation rules, and versioning (launched last month). This new capability adds another layer. Doyen flags the records that were impacted by a business rule entered by a human (humans can add business rules using natural language) and shows the user exactly what changed in each record.
How It Works
When a user writes an instruction like “Apply a 10% discount to all annual contracts,” Doyen translates it into executable mapping code.
The system then:
- Identifies which subset of records were impacted in the UI.
- Shows before-and-after values for each record.
- Allows users to validate and approve changes before committing them to their ERP or billing system.
Example: Billing System Migration
Instruction: “Reduce renewal price by 5% for customers who renewed after July 1.”
The system reveals:
- 482 subscription records impacted.
- Before: Renewal price = $1,000.
- After: Renewal price = $950.
Engineers and finance users can confirm that discounts applied only to eligible renewals and that no legacy invoices were unintentionally updated.
Example: ERP Migration
Instruction: “Reclassify all expense accounts starting with 6200 as Cost of Sales.”
The system shows:
- 213 journal lines updated.
- Before: Account 6205 – Marketing Expense.
- After: Account 5000 – Cost of Sales.
By exposing every affected GL entry, teams can verify that only the correct subset of transactions moved accounts.
AI-Human Interactions for Data Migrations
This feature is another step in our ongoing effort to make human–AI collaboration in data migrations as seamless and effective as possible. It’s part of creating the best possible experience for teams working alongside AI. More exciting features to come!