Where Procurement Loses Value — And How Agentic AI Can Stop the Leakage

Requirement Scoping, Post-Award Compliance, and Obligation Tracking: The Three Leakage Zones Where AI Impact Is Highest — and Human Stewardship Is Non-Negotiable.

Based on insights from the Forrester Opportunity Snapshot: “Don’t Delegate AI,” commissioned by Zycus, a survey of 261 procurement leaders (director-plus)

This series has examined agentic AI in procurement from multiple angles: who owns the strategy, where autonomy belongs, why execution lags vision, what foundations matter, how to govern without over-delegating, where to start with tail spend, and what actually holds scaling back. But every argument for deploying agentic AI ultimately rests on a single question: where does procurement actually lose value, and can AI recover it?

The Forrester study provides a precise answer. Value leakage concentrates in three areas — requirement scoping (48%), post-award compliance (48%), and obligation tracking (48%) — and leaders believe agentic AI can deliver its highest impact in exactly those domains: 75%, 69%, and 68%, respectively. This convergence of high leakage and high AI potential is the strongest business case in the study. It is also the domain that demands the most careful application of every governance principle this series has explored.

Requirement Scoping: Where Value Erodes Before Procurement Begins

Forty-eight percent of procurement leaders identify requirement scoping and scope creep as a top value leakage area. This is the upstream problem: before procurement can source, negotiate, or contract, the demand itself must be clearly defined. When requirements are ambiguous, incomplete, or shift after sourcing begins, every downstream activity inherits that imprecision. Suppliers bid on moving targets. Contracts fail to capture actual needs. And post-award disputes become inevitable.

Seventy-five percent of leaders believe agentic AI can have the highest impact here — the strongest endorsement in the study. AI can structure and validate demand at the point of intake, enforce completeness checks, flag ambiguities before they propagate, and benchmark requirements against historical patterns to catch scope creep before it becomes contractual.

This is where intake management becomes the critical control point. Zycus’s Merlin Agentic AI Platform addresses this through its Merlin Intake agent, which works within Microsoft Teams and Slack to capture demand in natural language, enforce policy compliance, and route requests to the right sourcing flow. Rather than relying on forms that business users fill out inconsistently, the AI structures every request from the first interaction — turning messy inputs into governed, actionable demand. As we argued in Part 4, cross-functional collaboration (a priority for 72% of leaders) starts at intake, and AI that governs the front door prevents leakage from ever entering the system.

Post-Award Compliance: Where Negotiated Value Disappears

The second leakage zone at 48% is post-award compliance. This is the midstream problem: procurement negotiates favorable terms, contracts are signed, and then the organization fails to enforce what was agreed. Pricing deviations go undetected. Volume commitments are not tracked. Rebates are unclaimed. The value that procurement fought for in negotiations quietly evaporates in execution.

Sixty-nine percent of leaders see high AI impact potential here. The opportunity is continuous, automated monitoring of contract compliance against actual transactions — something manual processes cannot sustain across hundreds of supplier relationships. AI can match invoices against contract terms in real time, flag pricing discrepancies as they occur, and trigger escalation before small deviations compound into material losses.

Zycus’s Merlin for Contracts provides the AI-powered contract intelligence needed here. It automates metadata extraction, identifies risk at each lifecycle stage, and surfaces compliance gaps that would otherwise go unnoticed until audit season. When combined with Zycus’s eProcurement solution — which monitors transactions in real time and flags policy violations automatically — the result is a closed-loop compliance system where negotiated value is actively protected rather than passively hoped for.

Obligation Tracking: Where Commitments Slip Through the Cracks

The third leakage area at 48% is tracking and managing obligations. Both procurement and suppliers carry commitments — delivery timelines, SLA performance, renewal dates, compliance certifications, and payment milestones — that must be continuously monitored. When obligations are tracked in spreadsheets or disconnected systems, deadlines are missed, penalties go unenforced, and renewal opportunities pass unnoticed.

Sixty-eight percent of leaders see high AI impact here. Agentic AI can continuously scan obligation data, alert teams to approaching deadlines, flag supplier non-performance against SLAs, and recommend actions before obligations lapse. This is governance at the operational level — the continuous monitoring that the three-pillar playbook from Part 5 was designed to enable.

Zycus’s Agentic AI for Supplier Management transforms obligation tracking from reactive spreadsheet reviews into autonomous orchestration. It synthesizes performance metrics, compliance data, and risk signals into a single control tower, continuously monitoring obligations across the entire supplier base and escalating exceptions before they become losses. For CPOs managing hundreds of supplier relationships, this is the difference between governance by exception and governance by design.

The Gray Zone Demands Procurement-Defined Rules

What makes these three leakage zones distinctive is that they sit in the gray zone between transactional and strategic work — the same territory we mapped in Part 2. They are too complex for pure automation but too voluminous for manual oversight. They need procurement-defined rules of engagement: explicit parameters for what gets flagged, what gets escalated, and what AI can resolve autonomously.

If these rules are set by IT or vendor configurations, organizations risk what the study warns against: using AI to accelerate fractured processes rather than correcting them. This is the central theme of Part 1 — CPOs must own these decisions because they carry the mandate, understand the workflows, and are accountable for outcomes.

Zycus’s AI-powered Spend Analysis underpins this governance by providing real-time, granular visibility. When spending data is continuously classified with up to 97% accuracy, the rules governing AI behavior are grounded in actual procurement patterns rather than assumptions. And Zycus’s Procurement Analytics agents make this data conversational, allowing leaders to query leakage patterns and refine rules in plain language.

Leakage Is the Cost of Not Leading

Value leakage across these three domains is not inevitable. It persists because they have historically been too labor-intensive to monitor at scale and too nuanced for blunt automation. Agentic AI changes that equation — but only when deployed under procurement-defined governance, with the right data foundation, orchestration layer, and balance of autonomy and oversight.

Every theme in this series converges here. Ownership (Part 1) determines who writes the rules. Autonomy calibration (Part 2) determines where AI acts independently. Readiness (Part 3) determines whether teams can execute. Foundations (Part 4) determine whether the data is trustworthy. Governance (Part 5) prevents over-delegation. Tail spend (Part 6) proved the model works. And reframing barriers (Part 7) showed the path is within reach. Value leakage is the ultimate test — and the ultimate reward — for procurement leaders who refuse to delegate the decisions that matter most.