New AI-driven procurement system aims to enforce existing rules more consistently at scale

Procurement has long been treated as a problem of rules. In many mid-to-large organisations, though, the rules are already in place: policy documents, approval matrices, vetted supplier lists and departmental budgets. The harder task is applying those controls consistently, every time, at the pace business teams expect.

That is the case made by Manjunath Hanmantgad in a build log for ProcureAI, a system he says is designed to automate the checks that should be deterministic whi...

le still leaving genuinely borderline decisions to people. The argument reflects a broader complaint seen across procurement software guides: Knack, Cflow and other workflow specialists note that manual approval chains tend to produce delays, inconsistent sign-off and poor visibility, particularly once request volumes rise. Procurify, meanwhile, warns that automation often fails when companies try to digitise a process that has not been properly defined in the first place.

Hanmantgad’s point is that inconsistency grows with scale. A manager might be able to spot exceptions when requests are few, but at higher volumes the same policy can be interpreted differently from one approval to the next. That creates practical problems as well as compliance risk. Decisions are then recorded in inboxes and spreadsheets rather than in a durable audit trail, leaving little evidence of why a request was approved, which clause justified it or what the budget position was at the time.

ProcureAI is presented as a way of breaking that pattern. Instead of relying on a single rule engine, the workflow is split into specialist agents that each handle one part of the evidence chain: policy checks, vendor status, budget verification, risk analysis and final approval assembly. According to Hanmantgad, the system is intended to return not just a decision, but the supporting material behind it. A blacklisted supplier, for example, would be rejected automatically; a request that exceeds budget would be declined with the remaining balance shown; and a compliant request would be routed to the right approvers based on the matrix already in force.

That emphasis on evidence mirrors a wider trend in procurement automation. Workflow platforms increasingly promise centralised routing, real-time tracking and better compliance controls because fragmented approval logic is hard to govern at scale. SysgenPro has argued that automation governance matters most where procurement connects to finance, audit and ERP systems, because weak orchestration in one area can undermine the whole approval chain.

Hanmantgad also acknowledges the limitations of his current build. The system is local-first rather than enterprise-integrated, its policy retrieval still depends on keyword search, and high-throughput use would require asynchronous processing. Human approver notifications have yet to be added, and the platform currently lacks authentication. Even so, the underlying thesis is clear: the real bottleneck in procurement is often not a missing rule, but an unreliable way of enforcing the rules that already exist.

Source: Noah Wire Services