AI-ready ERP starts with data the business can trust

AI-ready ERP is being marketed as a copilot sitting beside the human operator and quietly improving situational awareness. Acumatica’s forthcoming 2026 R2 practical-AI direction fits that narrative, bringing embedded intelligence closer to records.

However, AI-ready ERP is not chiefly about the interface; it is about whether an organisation’s data can survive contact with operational reality. In other words, while a copilot can spit out a confident answer in seconds, whether a user should act on it depends entirely on the records behind it.

An ERP is an organisation’s operational memory. It maintains records of customers, supplies, stock levels, invoices, production, and much more. Consider a situation where product names differ between warehouses, customer records are duplicated, lead times belong to last year, and critical decisions live in inboxes or spreadsheets rather than the system of record.

A copilot cannot repair that by sounding confident.

In the situations described above, asking which orders are at risk of missing their delivery dates or for stock levels would obviously lead to a wrong recommendation.

Embedded AI makes records easier to interrogate. It also makes their weaknesses consequential: an outdated assumption can quickly become advice that someone acts on.

Master data needs an owner

Businesses need named owners for customer, item, supplier and location records, agreed definitions and a clear process for correcting errors. Teams must be able to inspect the records behind a recommendation. Assistants must also respect access controls, particularly around sensitive margin and payroll information.

Start with read-only tasks: finding records, explaining exceptions and helping staff prioritise work. Check recommendations against actual outcomes and correct the source data where it falls short. Expand into actions only with defined limits, approvals and audit trails.

Acumatica is on the right track. 

Its emphasis on practical AI within existing workflows, with users retaining control, suggests it recognises these challenges. The next test is whether its tools can help employees identify when the available information is insufficient to support a decision.

As practical AI in ERP becomes an SMB battleground, buyers should ask vendors to demonstrate how their systems handle missing records, conflicting information and outdated assumptions.

The copilot demo is the easy part. 

The harder test is whether its advice holds up when the warehouse, purchasing team, and customer all have something different to say.

More ERP articles

• Interview with Claus Jepsen, Chief Technology Officer at Unit4
• Syspro Torque puts industrial AI to the test: Can auditable agents finally earn manufacturers’ trust?
• Why Syspro’s quality-first ERP creates a stronger foundation for industrial AI