Unlocking Generative AI: Transforming Procurement with Likwid AI

Generative AI in procurement: from pilots to governed, ERP-integrated workflows. Generative AI (GenAI) in procurement is quickly moving past “cool demo” territory. For manufacturing and distribution companies, the competitive

Generative AI in procurement: from pilots to governed, ERP-integrated workflows

Generative AI (GenAI) in procurement is quickly moving past “cool demo” territory. For manufacturing and distribution companies, the competitive edge in 2025–2026 will come from embedding GenAI into the real work of source-to-pay and plan-to-procure—grounded in trusted ERP master data (items, suppliers, contracts, inventory, and MRP signals) and wrapped in clear controls. That’s the difference between faster cycles with less value leakage and a new set of compliance and pricing risks.

The market signals are clear: interest is high, meaningful deployment is still limited, and the next wave of winners will be the organizations that treat GenAI as an operating-model change—not a chatbot experiment. EY reports that 80% of global CPOs plan to deploy GenAI in some capacity over the next three years, but only 36% have GenAI deployed “in a meaningful manner” (EY Global CPO Survey: 2025 Outlook).

Why GenAI hit the “trough of disillusionment”—and what manufacturers should do differently

GenAI for procurement is not failing; it’s maturing. Gartner noted that GenAI for procurement entered the “trough of disillusionment” in 2025, pointing to uneven ROI, fragmented or low-quality data, and integration complexity (Gartner press release (Jul 30, 2025)). In manufacturing and distribution, these issues show up fast because procurement is tightly coupled to planning, inventory, lead times, substitutions, and quality requirements.

The practical takeaway: GenAI outputs are only as reliable as the system context you give them. If your buyers and planners have to stitch together item specs from spreadsheets, supplier details from email threads, and contract terms from PDFs, GenAI will amplify the chaos. If GenAI is connected to a single platform’s governed data model and embedded in approvals and policies, it becomes a measurable productivity and compliance lever.

Highest-impact GenAI use cases inside manufacturing & distribution procurement

Not every process benefits equally. The best early wins are workflows with lots of text, repetitive decisions, and high cost of mistakes—especially where ERP data can “ground” the model.

Requisitions and guided buying (policy-aligned)

GenAI can convert natural-language requests (“need 500 meters of 304 stainless tube, 2-week lead time”) into structured requisitions by pulling from the item master, approved alternates, preferred suppliers, and approval thresholds. It can also explain why a request is being routed for approval (e.g., exceeds spend limit, non-preferred supplier, missing spec).

RFx creation and supplier communications

GenAI accelerates RFQ/RFP drafting using templates, commodity-specific requirements, and historical clauses—while ensuring each event includes correct specs, drawings, incoterms, and response fields. It can also summarize supplier Q&A threads into decisions that are easy to audit.

Contract management: summarization, risk flags, and obligation tracking

Real vendor progress is making this tangible. SAP highlighted an AI summarizer in SAP Ariba that can reduce document review times by 50% and described a “Sourcing Agent” (beta) intended to speed sourcing event creation (SAP Business AI release highlights (Q2 2025)). For ERP buyers, the key is not the summary itself—it’s linking the summary to contract metadata (effective dates, price breaks, MOQs, lead times, quality terms) so procurement and planning use the same truth.

Supplier onboarding and master data enrichment

GenAI can assist with onboarding by extracting key fields from W-9/W-8 forms, insurance certificates, ISO docs, and ESG questionnaires—then routing exceptions to the right reviewer. Done right, this improves data completeness and reduces downstream AP and receiving issues.

Invoice exception handling (3-way match) and dispute narratives

When invoices don’t match POs/receipts, GenAI can draft dispute emails, propose likely root causes (price variance vs. quantity vs. freight), and recommend next actions—grounded in PO terms, receipts, and contract pricing.

The data foundation GenAI needs to be reliable (and auditable)

If you want GenAI to write RFQs, recommend alternates, or flag contract risk, your ERP data has to be more than “available”—it must be governed and consistent.

At minimum, prioritize:

• Item master: specs, UOMs, approved alternates/substitutes, revision control, compliance attributes (RoHS/REACH), and preferred supplier links.
• Supplier master: site addresses, certifications, banking/payment details, lead-time performance history, risk flags, diversity/ESG attributes where relevant.
• Contract metadata: pricing terms, MOQs, incoterms, service levels, validity dates, rebates, and clause library references.
• Transactional history: PO/receipt/invoice lineage, price variance patterns, expedite history, and dispute outcomes.
• Planning signals: MRP recommendations, safety stock policies, inventory positions, demand changes, and critical component constraints.

Without these, GenAI tends to “sound confident” while making ungrounded suggestions. With them, GenAI can cite sources (e.g., “per contract C-1042 price tier 3 applies above 10,000 units”) and your team can verify quickly.

Preventing hallucinations and compliance risk: controls that actually work

Procurement isn’t a playground for free-form text generation. You need responsible-AI controls that fit regulated workflows, supplier relationships, and audit needs.

Grounding + citations + constrained actions

• Grounding: restrict answers to approved ERP data and vetted documents (contracts, policies, specs).
• Citations: require the assistant to reference the contract clause, PO, or supplier record used.
• Constrained actions: GenAI can draft, recommend, and pre-fill—but approvals, supplier selection, and contract execution should follow policy-based gates.

Human-in-the-loop for high-risk steps

Automate end-to-end only where mistakes are low impact and reversible (e.g., drafting a request email). Keep humans in the loop for supplier awards, contract redlines, and deviations from preferred sources or approved alternates.

ERP-native copilots/agents vs. bolt-on GenAI tools: integration and governance trade-offs

Bolt-on tools can be useful for narrow tasks (e.g., summarizing a PDF), but procurement value is created when GenAI is embedded where decisions happen: requisitions, sourcing events, PO creation, receiving, and invoice resolution. That requires deep integration with master data, roles/permissions, and audit trails.

It’s also where ROI separates leaders from followers. Deloitte reports that top-performing procurement organizations (“Digital Masters”) allocate up to 24% of their budgets to procurement technology (projected to rise to 26%) and achieve about a 3.2x return on GenAI investments versus slightly above 1.5x for followers (Deloitte 2025 Global CPO Survey press release). That kind of return typically comes from integrated workflows, not isolated tools.

Measuring GenAI ROI beyond headcount: the metrics that matter in manufacturing

GenAI business cases stall when they focus only on “time saved.” Manufacturers and distributors should measure outcomes tied to service levels, working capital, and compliance:

• Cycle time: requisition-to-PO, RFQ creation time, quote comparison time, invoice exception resolution.
• Compliance and leakage: contract price adherence, off-contract spend reduction, fewer maverick buys, fewer expediting fees.
• Planner productivity: faster response to shortages, better alternate sourcing decisions, fewer schedule disruptions.
• Supplier performance: on-time delivery, lead-time variability, quality incident correlation with lots/suppliers.
• Working capital: inventory turns, safety stock adjustments informed by lead-time risk signals.

McKinsey has emphasized GenAI’s potential to reduce procurement value leakage across compliance, contracting, and execution (Mitigating procurement value leakage with generative AI). That’s directly relevant to manufacturing, where a single missed clause or wrong spec can ripple into scrap, downtime, or customer penalties.

IP, confidentiality, and supplier data sharing: practical guardrails

Using GenAI for RFx drafting, contract summaries, and negotiation support raises legitimate concerns: pricing confidentiality, supplier IP, and controlled technical data. Build guardrails early:

• Data classification: tag drawings, specs, and pricing as restricted; block them from non-approved prompts and exports.
• Tenant isolation and retention controls: ensure prompts and outputs aren’t used for training outside your environment; log and retain interactions for audit.
• Supplier sharing rules: limit what the model can include in supplier-facing content; require a “send-ready” review step.
• Role-based access: buyers see what they’re allowed to see; engineering data and customer-specific requirements follow least-privilege access.

What “agentic AI” means in procurement—and what to automate safely

Agentic AI refers to GenAI systems that can take actions across steps—planning a task, gathering required information, generating outputs, and moving the workflow forward. In procurement, this can be powerful but must be bounded.

Good candidates for end-to-end (low risk)

• Drafting RFQ events from approved templates and item specs
• Summarizing supplier responses and highlighting deltas
• Generating internal justification drafts for approvals

Keep human-in-the-loop (high risk)

• Supplier award decisions and award communications
• Contract acceptance/redlines and deviations from standard terms
• Purchasing non-approved alternates for controlled parts

If you’re exploring this path, see additional perspective in The Benefits of Agentic AI in Procurement Processes.

GenAI connects procurement to planning and inventory—where manufacturers feel the biggest impact

Manufacturing procurement is inseparable from MRP-driven buying, substitutions, lead times, and constrained supply. GenAI becomes significantly more valuable when it can interpret planning signals and recommend actions that protect service levels:

• MRP-driven exception triage: explain why a part is critical (demand spike, delayed PO, scrap) and propose options.
• Alternate sourcing suggestions: recommend approved alternates and validated suppliers when lead-time risk spikes.
• Lead-time risk detection: flag suppliers with worsening variance and suggest mitigation (split buys, safety stock changes).

This is why ERP buyers should care about a unified platform. If GenAI can’t see the same MRP and inventory truth your planners use, it can’t reliably reduce shortages or expedite costs. For more on connecting AI to planning, explore AI-Native MRP.

Rollout plan: how to deploy GenAI without breaking procurement controls

Adoption requires more than enabling a feature. A practical rollout sequence looks like this:

• Start with 2–3 pilot use cases (e.g., RFx drafting, contract summarization, invoice exceptions) and define success metrics.
• Fix the data bottlenecks that would cause unreliable outputs (item and supplier master completeness, contract metadata).
• Implement governance: role-based access, audit logs, prompt/output retention, and approval gates aligned to policy.
• Train buyers and planners on what the tool can/can’t do and how to validate outputs.
• Scale by workflow, not by “number of users,” expanding into requisitions, guided buying, and plan-to-procure exceptions.

If you’re comparing approaches, AI vs Traditional Procurement: Which is Better for You? and 10 Procurement Automation Workflows to Cut PO Cycle Time provide helpful frameworks for prioritization.

How Likwid AI helps manufacturers operationalize GenAI in procurement

Likwid AI is built for manufacturing and distribution complexity—so you can manage procurement, production planning, inventory. If you’re ready to move from experiments to governed workflows, explore Procurement AI and see how Likwid AI can help you turn GenAI into measurable procurement performance.