Digital Transformation Methods

Transforming a Traditional Business Without Replacing Your ERP All at Once: Diagnose, Digitize, Then Bring in AI

Systems nobody uses, training that changes nothing, ERP migrations that stall operations: most of these come from doing things in the wrong order. Here is how Twouring approaches it in three steps, what each step delivers, how it is accepted, and a short checklist to find your starting point.

When we talk with owners of traditional businesses about digital transformation, the first question is usually: "Should we replace our ERP?" We usually hold off on answering. Whether to change systems depends on how work actually happens on the floor. In many companies, the real bottlenecks are quotes that live in one person's head, inventory numbers that never match, and production status that requires a phone call. None of these necessarily start with a new system.

01Three common failures, and why order is usually the problem

The first is the system nobody uses. Nobody mapped how sales, production control and the warehouse actually work before rollout, so the fields do not match real tasks. Staff end up entering data into the system while still maintaining their own Excel files. After a while nobody trusts the system data, and everyone goes back to the old way.

The second is sending people to training and seeing nothing change. External AI or software courses use generic examples that have nothing to do with your quotation forms, part-number rules or customer inquiry formats. People understand the lesson, then return to their desks without knowing where to start tomorrow morning.

The third is replacing the entire ERP in one go. During cutover, orders, inventory and accounting all change at once, and a snag in any one of them hits shipping and collections directly. For small and mid-sized companies that are already short-handed, this period is often riskier than expected.

All three share one thing: the tool was chosen before anyone looked at the process. We do it the other way around. We get a clear picture of the process first, then decide what belongs in a system and what is a good fit for AI.

02Step 1: Diagnosis, finding the three most labor-intensive steps

The diagnosis takes about two to three weeks. Our consultants go on site and interview the people responsible for quoting, order intake, production control, the warehouse and accounting. We follow one order from inquiry to shipment and payment, counting how many people, forms and repeated data entries it passes through.

The deliverable is a diagnostic report listing the three steps that consume the most labor and produce the most errors, with priorities, a suggested timeline and a budget range. Acceptance is simple: after reading it, do the owner and department heads agree these three are the most painful points today? If people disagree on the ranking, we go back for more interviews until there is consensus.

The diagnosis can stand alone. You can decide whether to continue after seeing the results, or take the report to your existing system vendor. We once helped a research institution map its data workflows. The conclusion was to hand standardizable sorting work to the system and keep judgment calls with people. You only see that kind of split once the process is laid out.

03Step 2: Digitizing the process, integrating where possible instead of forcing replacement

Step 2 tackles the steps the diagnosis identified, moving quotes, orders, part numbers, inventory and production progress out of Excel and paper and into a system. If you already have an ERP whose core functions still work, we prioritize integration so new tools feed data into the existing system. We do not require you to replace everything.

The deliverable is a process and data set that runs every day: part numbers follow one coding rule, orders move through fixed statuses from creation to shipment, and inventory has a single source of truth. Acceptance is judged on daily work. Can sales look up an order's production status directly in the system? Does the warehouse count match the system? Does month-end reconciliation still require comparing several spreadsheets by hand?

One hardware and materials company carried thousands of part numbers across e-commerce, an ERP and physical retail counters. Our entry point was product data: first consolidate part numbers, specifications and prices into one shared master file, then have e-commerce, the ERP and the counters all pull from it. Once product data is in order, connecting any later system becomes much easier.

04Step 3: AI adoption and in-house training tied to each department's daily work

AI needs data to work with, which is why it comes after digitization. In Step 3 we design AI applications around the company's actual work, such as organizing quotations, replying to customer inquiries, producing export documents, writing meeting minutes and compiling management reports.

Then we run hands-on classes by department. Each class practices what that department does every day: sales practices inquiry replies, shipping practices export documents, managers practice report compilation, so people can use it as soon as they return to their desks. Acceptance looks at actual use after class, meaning whether the designated tasks are now done the new way.

We once designed a 24-hour AI workplace training program for hospital medical staff: eight topic sessions, run in cohorts by four job categories, with simulated data used in every class. Training for traditional businesses follows the same principle. People are grouped by role and practice with simulated data, and no real customer or company data is brought into the classroom.

05Where should you start? A self-check list

Not every company needs to start at Step 1. Ask yourself these four questions:

1. You cannot name the three most labor-intensive steps in your company, or department heads give different answers. If so, start with diagnosis.

2. Quotes, orders or inventory still live mainly in Excel, on paper or in LINE chats, and the same data is entered two or more times. If so, you can start with process digitization, but do a quick review before work begins.

3. Your core data is already in a system, searchable and consistent, yet staff still spend a lot of time preparing documents, replying to inquiries and building reports. If so, you can go straight to AI adoption and training.

4. You have rolled out a system or sent people to training before, but it did not stick. In that case, go back to diagnosis and find out what got stuck before deciding the next step.

06Three mistakes to avoid

First, picking tools without a diagnosis. However complete a vendor demo looks, it will not get used if it does not fit your process.

Second, changing too many departments at once. Pick one step, make a visible change that staff can feel, then expand to the next department. Resistance will be much lower.

Third, handing transformation to younger staff or IT alone. It changes how every department works, so the owner or successor needs to take part in setting priorities. Otherwise each department tends to go its own way. If you are not sure where to start, join our online briefing, where we walk through the three steps using real workflows.

Portrait of Kai Hsu (KK)

Author

Kai Hsu (KK)

Founder & lead advisor, Twouring Co., Ltd.

LinkedIn →

Start with a 30-minute discovery call to confirm your next step

The call is free. We confirm the scope of the problem, the budget and the decision-maker, then give you initial recommendations and a practical next step.