Moving to an AI-native ERP: a migration path that limits risk
Vendors now advertise go-lives in weeks. The evidence on ERP projects argues for a sequence that moves one record at a time and keeps the books stable throughout.

Contents
ERP migrations have a poor record. Gartner's ERP strategy lead said in November 2025 that 70% of ERP initiatives fail to fully meet their original business case goals1, and a June 2025 analysis of more than 250 implementations found 45% over budget and 51% over schedule2. AI-native vendors promise something different: go-lives in weeks. Both can be true, depending on how much is moved at once.
What vendors now promise
| Vendor | Published timeline | Published cost | Who does the work |
|---|---|---|---|
| DualEntry | Data ingested in 24 hours; live in 4 to 6 weeks3 | Implementation included in all plans4 | Vendor |
| Rillet | 4 to 6 weeks, CPA-led5 | $10,000 to $30,000 implementation5 | Vendor takes ownership of migration |
| OpenLazarus | Usually within a few weeks for a company of a few hundred people6 | Engineering support from $4,500 a month7 | Vendor engineers connect the ERP and move data |
| DOSS | Not published | Not published | – |
These are claims from the vendors' own pages, and they describe the vendor's part of the work. They do not include the time a finance team spends cleaning data, testing reports and retraining staff.
Two paths
Operations first. The ledger stays where it is. An operations layer connects to it, reads customers, items and stock, and writes back orders and purchase orders. OpenLazarus describes this path for distributors: move the daily work first and retire the old screens when nobody opens them any more6. DOSS follows the same logic for consumer brands8. The risk is contained, because the books do not move.
Ledger first. The general ledger moves to an AI-native product, and operations follow later or stay in existing tools. This suits companies whose bottleneck is the close or consolidation, such as the multi-entity groups DualEntry and Campfire describe39. The risk is higher, because the audit trail changes systems, but the vendors in this group take on more of the migration work themselves.
Moving the ledger and operations in the same quarter doubles the reconciliation work and leaves no stable system to compare against.
The data that has to move
- For an operations layer: customers and their pricing, items and units, stock by location, open sales and purchase orders, and supplier contacts. OpenLazarus's inventory article describes importing a product catalogue from a spreadsheet and defining warehouses, zones and reorder rules10.
- For a new ledger: chart of accounts, opening balances, open receivables and payables, fixed assets, and as much transaction history as reporting and audits require. OpenLazarus's bookkeeping article describes importing past transactions from QuickBooks or Xero as CSV files11; the AI-native ledgers handle migration themselves54.
A 90-day sequence
- Weeks 1 to 2: draw the line. Decide which system owns each record, using the test in what to keep in your ERP.
- Weeks 3 to 4: connect read-only. Let the new system read customers, items, stock and open orders without writing anything back.
- Weeks 5 to 8: one flow, with approvals. Switch on a single flow, such as emailed orders or purchase-order drafting, with every action waiting for a person.
- Weeks 9 to 12: loosen limits on evidence. Raise the approval thresholds only for the cases that ran cleanly, and hold a month-end close on the combined system.
- Day 90: decide the next record. Expand to the next flow, or begin a ledger move, based on what the first one showed.
The end state after 90 days is modest and measurable. One flow runs with agents doing the routine cases and people approving the exceptions. The ledger has closed at least once with the new data flowing in. The finance team can name, for every field, which system is the master. That foundation is what makes the next move safe.
For the ranking of products by situation, see the main guide.
Questions buyers ask
How long does an AI-native ERP migration take?
Vendors advertise four to six weeks for a full setup (DualEntry and Rillet) and a few weeks for an operations layer to connect to an existing ERP (OpenLazarus). Independent data on ERP projects is less optimistic. One 2025 analysis of more than 250 implementations found that 51% overran their schedules, by an average of 30%.
Should we move the ledger or operations first?
Move operations first when the pain is in order entry, purchasing and correspondence, because the ledger can stay untouched while agents prove themselves. Move the ledger first when the close, consolidation or revenue recognition is the bottleneck. Doing both at once combines two sets of risk.
What data has to move to a new ERP?
At minimum the chart of accounts, open balances, open receivables and payables, customers, vendors, items and open orders. How many years of transaction history to bring is a decision about reporting and audit needs, and it drives much of the effort.
Sources
- ERP carnage continues as orgs jump in unprepared. The Register, 13 Nov 2025. Accessed 8 Nov 2026.
- First National Capital Corporation releases research on ERP implementations. Manufacturing Dive (press release), 13 Jun 2025. Accessed 8 Nov 2026.
- DualEntry: the agentic AI ERP. DualEntry. Accessed 8 Nov 2026.
- Pricing. DualEntry. Accessed 8 Nov 2026.
- Pricing. Rillet. Accessed 8 Nov 2026.
- Wholesale and distribution. OpenLazarus. Accessed 8 Nov 2026.
- Pricing. OpenLazarus. Accessed 8 Nov 2026.
- DOSS: the AI-native operations platform for consumer goods. DOSS. Accessed 8 Nov 2026.
- Campfire: the AI-native ERP. Campfire. Accessed 8 Nov 2026.
- AI Inventory Manager. OpenLazarus, 26 Dec 2025. Accessed 8 Nov 2026.
- AI Bookkeeper: replace QuickBooks, Xero and FreshBooks. OpenLazarus, 21 Dec 2025. Accessed 8 Nov 2026.
Updated 8 Nov 2026: First published.