Epicor launches AI agents in its logistics ERP — the market's conviction about the warehouse is confirmed, but the question that matters stays the same
On 6 August, Epicor — one of the major supply-chain software vendors — launched "Prism", AI agents built directly into its ERP suite, first in Australia and New Zealand. Futurum, an analyst firm that tracks the sector, calls the launch a "major shift" for AI in logistics. The same day, at AppH, Warehouse and Automations each received a real new capability — the same conviction, at a completely different scale.
Prism joins Epicor's ERP suite to help mid-sized manufacturers and distributors automate inventory planning, forecasting and repetitive supply-chain tasks. The rollout starts in Australia and New Zealand before a wider expansion. Futurum is not a press release: it is an analyst firm that follows the sector closely, and its verdict — "major shift" — lines up with two signals already cited in our previous digests: the Houlihan Lokey report of 23 July, which named fleet and tourism among the verticals with the strongest conviction for embedded-AI investment, and the Upwork SMB survey, which placed inventory management among only 3 business functions that have genuinely moved past the pilot stage.
On the same day as that launch, at AppH, two concrete things happened, not one: Warehouse received its stock-expiry tracking yesterday, and today the Automations engine gained the ability for the owner to define THEIR OWN rules — a condition on a real field (stock, due date, status) without writing a line of code. What neither that digest nor Epicor's product page specifies for Prism, and what we can verify in our own code: none of the Automations rules — fixed or custom — executes an irreversible action on its own. The purchase-order draft that low stock triggers stays a draft until a human validates it; an expiry alert opens an event to be handled, it neither archives nor orders anything by itself.
For AppH
- This is not a roadmap promise: the custom-rules feature is real code, tested (+6 tests), deployed and verifiable as of today — not a launch announced for "soon".
- Every rule, fixed or custom, only opens an event for a human to handle — never an order, an email or a financial decision executed on its own. That is verifiable in the code, not just asserted in this article.
Against / the honest limit
- We do not know whether Prism's agents distinguish "recommend" from "execute" — Epicor does not say so publicly at this stage, and "major shift" is the analyst's judgement, not an independent check we were able to make ourselves.
- AppH's custom-rules engine covers only 3 entities today (leads, stock, tasks) with simple conditions (number, text, date) — far from a full workflow builder. It is a real start, not feature parity with an entire ERP suite.
The temptation would be to write that this launch proves us right — but the real comparison is not a feature race with an ERP vendor far larger than AppH. It is a design question we prefer to state plainly rather than leave implicit: an AI that recommends an action is not the same as an AI that executes it, and the enterprise-AI market has an unfortunate habit of blurring that line in its announcements. At AppH, the rule does not change depending on whether the condition comes from one of the 80 built-in templates or from a rule the owner wrote themselves five minutes ago: no action with real consequences — an order, an email, a financial decision — leaves without a human validating it first.
Verified by a human at AppH