Skip to content

When there is no usable ERP

Our default is to attach to the ERP you already run. This page covers the other case: there is no ERP worth attaching to, so our forward-deployed engineers build an ERP-like core as part of the operations layer, and you own it.

Three situations count as no usable ERP.

  • Missing. The books are in an accounting package, and orders, inventory and purchasing live in Excel, email and people’s heads.
  • At end of life. The vendor no longer supports the system, it runs on hardware nobody wants to restart or the one person who can change it is about to retire.
  • Cannot be extended. There is no supported way to get data in or out, so nothing can be built around it.

An ERP that is disliked, slow to change or poorly set up is still usable, and we attach to it. The call is made on evidence. The Systems and Data Audit from the Discovery Sprint records, for each system, how to get data in and out. If the answer is that there is no supportable way, we tell you at the readout and show you why. Whether to build a core is then your decision.

The core covers four areas, and in each one only what your company uses.

Area What it holds
Orders Customers, sales orders and shipments
Inventory Items, stock locations, quantities on hand and every movement
Purchasing Suppliers, purchase orders and receipts
Invoicing Customer invoices and credits

“Only what your company uses” is a working rule. If you ship from one warehouse, the core has one warehouse. If you do not track lots, it has no lot tracking. Each thing the core does is there because our engineers watched someone do that work on your floor or in your front office. More can be added when the work changes.

The core is not a general-purpose ERP and not a packaged product. It has no general ledger. Your accounting system stays the system of record for the books, and the core passes invoices to it the same way the operations layer attaches to any other core system.

The core is built as part of the operations layer, and it takes the place an ERP would have: it becomes the source of truth for orders, inventory, purchasing and invoicing.

We hold it to the same rule we apply to an ERP. Screens and agents change its records only through its own checked transactions, never by editing its tables, and every change is recorded with who made it. Agents work on the core as they would on an ERP: from queues, inside limits, under the autonomy ladder.

Moving off the old system and the spreadsheets happens in three moves. There is no single weekend where everything changes at once.

  1. Mirror the old data. The core first holds a read-only copy of the data in the old system and in the spreadsheets that matter, refreshed on a schedule. The work still happens in the old place. Your people see their own customers, items and open orders in the new screens, and the data problems show up early: duplicate customers, dead items, three spellings of the same unit of measure. The people who know the data decide how to clean it.
  2. Run in parallel. One workflow at a time, the work is done in the new screens and in the old system side by side. The results are compared: the same orders, the same quantities on hand, the same invoice totals. Every difference is traced to its cause. The parallel run for a workflow ends when the comparison has stayed clean for a period agreed in advance with the workflow owner.
  3. Cut over workflow by workflow. When the workflow owner signs off, the core becomes the source of truth for that workflow and the old system stops being updated for it. The next workflow follows when it is ready. The old system is retired when the last one has moved, and its data is kept as a read-only archive.

This way of migrating has two costs, and you should know them before you start.

  • During a parallel run, some work is done twice. That is the price of not betting the company on one weekend.
  • While some workflows run on the core and others still run on the old system, the two have to exchange data so that each sees the other’s transactions. That bridge is temporary work, and it is one reason the order of the cutovers matters.

This is the approach in principle. The order, the timing and the length of each parallel run are planned with you for your company.

The core’s code and every record in it are yours, on the same terms as everything else we build for you: you own the custom code and the data, and we keep our reusable tooling. The core is written for you. The parts around it that come from our reusable tooling are the work queue, the approval screens, the audit trail and the stop switch, and your contract gives you the right to keep running and changing that copy after we leave, with no license fee. Where the core is hosted is agreed in the contract. Stages and terms lists what your contract records.

Owning it has a consequence worth saying plainly. No vendor ships upgrades for a core built for one company. It changes when you ask for a change, and someone has to run it: us under Operate & Automate, your own people or another vendor you choose. Handover and exit covers how a handover works.