M&C ERP and AI: why clean data is the condition for AI to be truly useful

We’re at a point where nearly every enterprise technology conversation ends with the same question: “So what can AI do?” For import businesses — like all other businesses — many providers’ answer is to attach an AI feature to their product. This article wants to say the opposite: for import trade businesses, the AI story doesn’t start with language models, but with ERP data quality. Before deploying AI, you need trustworthy data underneath.

Numbers worth considering

The dirty data problem isn’t subjective — it has numbers:

• Gartner estimates that poor data quality costs each organization an average of $12.9 million per year [1].
• Same Gartner: about 60% of AI projects are expected to be abandoned by 2026 — the leading cause is data problems, not technology [2].
• IBM once estimated that bad data costs the US economy up to $3.1 trillion annually [1].

Explained briefly by the classic garbage in, garbage out principle: deploy AI on fragmented, contradictory, non-standard data — you get fluent, confident, and wrong answers. That’s not artificial intelligence; that’s a faster way to make wrong decisions. Fujigo analyzed this theme in AI strategy needs a solid ERP foundation — this article specializes for the import trade industry.

Why is the import industry particularly sensitive to dirty data?

Import business from Japan is a natural “data generator”: one shipment brings along purchase orders (JPY), bills of lading, customs declarations, goods receipts, invoices — each document a set of numbers, each department a copy. If these copies are in multiple systems, “this number matches that number” becomes manual human work — and like all manual work, it drifts over time.

Result: when you need a number — the real cost of an item, available inventory, accounts receivable balance — someone has to manually combine. Manual combination is slow, and slow means the numbers used for decision-making are always “yesterday’s” numbers.

ERP first, AI later: standard data is the condition

M&C ERP is designed on the principle of each document entered once, reused later: purchase orders generate shipments, declarations feed cost, serials are generated in the warehouse for delivery and warranty (detailed document cycle in A document cycle in M&C ERP). When the chain runs correctly, the system has what AI needs more than large models: a single source of truth — correct cost, correct inventory, correct receivables, interconnected.

This is a necessary condition. Without it, every question to AI is answered on a wrong data foundation.

So what does AI do on a clean data foundation?

When ERP data is standard, AI applications become practical scenarios instead of showy — for the import industry, the natural direction is:

• Search and Q&A: “Where is the shipment from supplier X, when does it arrive?” — instead of opening three screens and an Excel file.
• Aggregation and reporting: summarize inventory, receivables, import costs by period — already in the system, now needing a faster way to present.
• Alerts: supplementary filing deadline reminders, inventory hitting thresholds, unreconciled invoices — the system proactively calls out instead of users having to remember.

Important note: these are directions based on the foundation, not feature list commitments. The correct order remains: business standardization first — cost matches, documents match, inventory matches — then AI has a place to stand.

Conclusion: AI is the reward of data discipline

For Japanese businesses investing in Vietnam and import companies, the practical message is: if today you’re tracking shipments by email and cost by Excel, then the most valuable investment step for the AI era isn’t buying another AI tool — it’s bringing business data into a standard system. AI will come later, and when it does, it will have real work to do.

---

Want to see how that data foundation forms in a specialized ERP system? Contact us for a demo account of M&C ERP.

---

References

1. IBM — “What Is Data Quality?”: Gartner estimates poor data quality costs each organization an average of $12.9 million/year; garbage in, garbage out principle in AI/ML. https://www.ibm.com/think/topics/data-quality
2. VexData (citing Gartner 2026) — 60% of AI projects abandoned by 2026 due to data problems; 63% of organizations lack AI-ready data practices. https://www.vexdata.io/post/garbage-in-garbage-out-the-business-case-for-data-quality-in-2026