Trending Topics

AI can build ERP customisations 60% faster. Can change control keep up?
Imagine renovating a building while people are still working on living inside it. Even a small alteration requires planning, skilled labour and checks to ensure residents are not disrupted. Changing an ERP system has traditionally been like that – customisations have normally taken time to design, build and test. But AI is now speeding this up.
Epicor says Prism Developer can reduce the time required to build and test Application Studio customisations by an average of 60%. Faster development sounds like an obvious win for manufacturers and distributors waiting on workflow changes.
But coding was never the only bottleneck.
A faster route to change
Epicor expanded Prism’s availability across Australia and New Zealand in August. The company says its embedded ERP offering includes more than 18 pre-built AI agents. For Kinetic customers, Prism Developer uses natural-language prompts to generate Application Studio components, with changes previewed and validated before they are merged.
Epicor’s claimed build-and-test time reduction, illustrated against a 100-unit baseline:
| Without the claimed saving | With the claimed saving |
| 100 units of time | 40 units of time |
The illustration does not show that approvals, deployment, or the entire ERP project become 60% faster.
That distinction is critical.
If teams can produce more proposed changes in less time, reviewers may face more work, not less. A useful screen adjustment could sit alongside an undocumented rule that changes purchasing approvals or gives a role access to data it should not see.
Governance becomes the bottleneck
Every AI-generated customisation still needs an owner, a reason for existing, testing against dependent processes, and a rollback plan. Role-based access and segregation of duties cannot be treated as paperwork to complete after deployment.
Our examination of Epicor Prism and Infor shows how ERP vendors are moving towards AI-assisted work inside existing platforms. But, as the analysis of process mining and AI agents argues, automating a poorly understood process can compound its problems.
Prism Developer could remove tedious development work. The test is whether organisations invest some of the time saved in reviewing, documenting and governing the changes it helps create.
