AI in Procurement Use Cases

Jul 2, 2025 • Zeiv

With all the noise around AI in procurement, I set out to cut through the hype - especially the bold claims made by AI evangelists - and understand the ground reality.

This is an ongoing series where I’ll break down one use case at a time. Each section looks at the major claims made by AI advocates, what’s actually happening on the ground, and how procurement teams can realistically leverage AI to improve their processes.

So far, this article covers:

AI in Supplier Discovery

The Claim

If you look at most procurement tech evangelists, here’s what they claim:

  • AI can identify qualified suppliers from global databases using specs, industry tags, or past spend data.

  • It can score and rank suppliers based on risk, ESG, diversity, financials, or delivery performance.

  • Generative AI can even draft RFIs or outreach emails to potential suppliers.

  • Some claim “autonomous sourcing”, where AI finds, invites, and helps evaluate suppliers without human input.

Sounds slick on paper.

The Reality

  • Most tools are only as good as their data coverage. If the supplier isn’t in the network, AI won’t magically discover them.

  • Matching is often surface-level: based on keywords, product categories, or UNSPSC codes. No understanding of fit for niche requirements.

  • Supplier scoring is usually backward-looking, based on generic risk feeds or limited history. It doesn’t know your unique business needs or current supplier performance nuance.

  • Outreach is template-based. It saves time but doesn’t replace thoughtful engagement that builds trust.

In short: AI can help narrow the pool, but it doesn’t know who’s worth betting on.

What It Lacks

  •  Contextual judgment: It doesn’t know that a certain supplier is “good on paper” but flaky in execution.

  • Local insights: A seasoned buyer knows who’s really doing the work behind a supplier name, or who to avoid in a particular region.

  • Fit-for-purpose nuance: AI might find a supplier who makes a part, but not whether they can scale, customize, or deliver to your lead time needs.

  • Relationship capital: You can’t replace 10 years of working together, knowing which vendor will go out of their way to fix a late order.

So Where Does AI Actually Help?

Initial Search Acceleration

AI can process massive volumes of structured and unstructured data to surface possible suppliers faster than a human could. It can scrap supplier directories, websites, past RFx data, and even documents to build a candidate list. Especially useful for unfamiliar geographies or categories where the buyer lacks experience.

Filtering and Categorization

AI can sort suppliers by attributes: certifications, product specs, region, size, past spend, diversity tags, etc. Think of it like a turbocharged search engine for supplier databases and it's good for narrowing a large pool into a short list that a human can evaluate.

Pattern Recognition Across Spend Data

AI can analyze internal spend and show you: Suppliers used by other departments that you didn’t know about, missed consolidation opportunities, and fragmentation across similar purchases.

Automating Low-Stakes Outreach

For non-critical, tail-spend categories, AI can auto-generate RFIs, emails, or supplier forms. Saves time for repetitive tasks, where you don’t need deep engagement.

How Buyers Can Benefit (Without Getting Burned)

Treat AI as a Scout, Not a Selector

Use it to find options, not to make decisions. AI is good at surfacing leads, you still need to vet, validate, and sense-check.

Use It to Cover Blind Spots

AI can show you suppliers you’d never find manually, especially in fragmented or new markets. That’s where it truly adds value.

Don’t Assume Scores Are Truth

If a tool ranks a supplier as “high risk” or “top choice,” understand how that score is generated. Use it as a signal, not a fact.

Plug It Into Your Process, Not Over It

AI should fit into your sourcing workflow, not replace it. Let it reduce grunt work, but keep human evaluation at the center.

Bottom line

 AI is a helpful tool for surfacing supplier options and saving time, especially when starting from zero. But it doesn’t replace buyer experience, relationships, or judgment. It’s a co-pilot, not a captain.

AI in Supplier Evaluation

The Claim

Procurement tech platforms promise that AI can:

  • Score suppliers based on performance metrics, delivery data, or compliance records.

  • Pull in external risk signals (e.g. financial risk, ESG scores, sanctions, news sentiment).

  • Highlight preferred suppliers using predictive models.

  • Enable "data-driven selection" of suppliers—removing bias and manual guesswork.

Some even claim real-time risk alerts and automatic adjustments to sourcing decisions.

The Reality

Let’s be blunt: AI can process and surface risk indicators. But it doesn’t understand context the way a buyer does.

  • Most scoring models are heavily dependent on the quality and availability of internal (your ERP history) and external (third-party feeds) data.

  • False positives and missed signals are common. A supplier might get flagged due to generic negative news that has no impact on your category.

  • In most organizations performance data is incomplete especially for newer or smaller vendors. So, no data = no score.

  • AI doesn’t account for relationship dynamics, workarounds, or how a supplier has stepped up during tough times.

It’s helpful. But it lacks the full picture to operate autonomously.

What It Lacks

  • AI lacks granular context. For example: AI can’t distinguish between a late delivery due to port delays vs. poor planning.

  • AI models tend to flag suppliers based on universal risk indicators like financial health, credit score, or negative news mentions. But, in reality what looks like a red flag in one category may be an acceptable or manageable trade-off in another.

  • AI also lacks buyer judgment. It doesn’t know when to make an exception or push a supplier to improve rather than replace them.

Where AI Actually Helps

1. Consolidating Risk Signals: AI can surface relevant info fast from Dun & Bradstreet scores to sanction lists and negative media coverage.

2. Flagging Anomalies: If a supplier’s delivery score drops, payment terms shift, or pricing becomes erratic, AI can highlight it early.

3. Comparative Scoring (at Scale): If you're evaluating 20+ vendors across standard metrics (on-time delivery %, complaint rate, cost overrun), AI can stack-rank consistently.

4. Reducing Manual Reviews: For repeat purchases or low-risk categories, AI can help fast-track routine evaluation with automated checks.

How Buyers Can Leverage It

  • Use AI to flag, not finalize decisions. Dig deeper when the score looks off and take a judgement call.

  • Supplement AI with supplier feedback, internal stakeholder input, and category nuance.

  • Revisit how supplier performance data is collected. AI is only as good as what you feed it.

Bottom Line

AI can surface patterns, flag risks, and speed up supplier evaluation. But it doesn’t replace buyer judgment. Without context, nuance, and good data, even the smartest model can mislead/hallucinate.

AI in Price Benchmarking

The Claim

Procurement tech vendors claim AI can:

  • Benchmark supplier prices in real time using internal PO data and external market feeds.

  • Predict the “should-cost” of a product or service based on components, labor, geography, and time.

  • Flag outliers and overpricing before you even look at the quote.

  • Suggest negotiation ranges or recommended price adjustments.

  • Run continuous market comparisons for high-volume or repeat purchases.

In theory, this means no more manual Excel benchmarking or chasing reference quotes and you can just get instant cost intelligence.

The Reality

AI can help, but only when there’s enough clean, structured data behind the scenes. Here's where it often falls short:

  • Data fragmentation: Internal PO data is often messy, unstandardized, or scattered across systems. Garbage in = garbage benchmark.

  • External data gaps: Market feeds are limited by geography, currency, and category. Custom products or services rarely have clean benchmarks.

  • Context blindness: AI might flag a price as “high,” but not realize it's due to premium lead times, bundled services, or inflation-adjusted rates.

  • Limited adaptability in dynamic categories: For services, creative work, or one-off builds, AI can't predict true cost without understanding specs or outcomes.

So yes, it can highlight potential red flags but only under the right conditions.

What It Lacks

  • Spec-based nuance: AI doesn’t always account for variation in quality, grade, delivery terms, or packaging that justify price differences.

  • Historical context: A sudden price increase might reflect a real market shift and not supplier margin padding. AI doesn’t know unless you feed it that information.

  • Cost drivers outside the data: Tariffs, supplier risk premiums, or internal process changes are hard for AI to factor into pricing models.

  • Negotiation leverage: AI can’t yet weigh strategic factors like urgency, supplier dependency, or relationship capital when assessing “fair price.”

Where AI Actually Helps

  • Identifying price outliers: AI can scan historical spend and flag when a new quote is significantly above or below previous purchases for similar items.
  • Spend trend analysis: It helps to visualize how costs have evolved over time by category, supplier, or region. It is useful in budget planning or category strategy.
  • Should-cost modeling (in structured categories): For manufacturing or commodity-heavy purchases, AI can estimate base material + labor + overhead to guide negotiations.
  • Tail-spend benchmarking: AI is effective in flagging inconsistent pricing across small-ticket or repeat purchases, places where humans rarely check.

How Buyers Can Leverage It

  • Use AI as a sanity check, not a pricing authority. Let it surface price anomalies and ranges but you still drive the context and decision.

  • Pair AI insight with specs, supplier input, and market news before making assumptions.

  • Use it to guide negotiation prep, not automate negotiation itself.

  • Focus on cleaning and tagging internal pricing data that’s what enables accurate benchmarking in the first place.

Bottom Line

AI helps you spend smarter, not just faster. It speeds up benchmarking, flags price risks, and gives buyers a head start. But it can’t replace judgment, market knowledge, or real-time context. Think of it as your pricing radar, not your pricing oracle.

To be continued....