Modern Category Management: 5 AI-Ready Best Practices

Most procurement teams still treat category management like a once-a-year planning exercise: refresh the Kraljic matrix, negotiate a few big contracts, publish a slide deck, repeat. That’s not category management anymore—it’s a calendar ritual.

Category management best practice in 2025 looks closer to a living operating system: it updates when demand shifts, when suppliers wobble, when regulations change, and when your business decides (again) that resilience matters more than unit cost. Spend Matters’ recent research (including interviews with senior leaders such as Sanofi’s Country Procurement Head, Fabio Fontes) lands on a blunt truth: CatMan is still central—but the format has changed.

This post is part of our “AI in Supply Chain & Procurement” series, so I’ll take those classic CatMan fundamentals and show where AI actually helps—and where it doesn’t. You’ll get five practical insights you can apply to your next category strategy refresh, plus a simple blueprint to make your category plans more adaptive without turning them into science projects.

1) Category management still matters—your format is what’s outdated

Answer first: Category management remains the core discipline for aligning spend, suppliers, and business outcomes, but static frameworks fail in volatile markets.

Kraljic’s matrix (introduced in the late 1970s and popularized in the early 1980s) taught procurement leaders to segment categories by profit impact and supply risk. That logic still holds. What doesn’t hold is pretending those axes are stable.

In 2025, “risk” changes fast:

• A single quality incident can shut down a biopharma supply lane.
• Geopolitical developments can reroute logistics overnight.
• ESG and human-rights scrutiny can turn a low-cost supplier into a board-level liability.

The practical shift is this: your category strategy shouldn’t be a document; it should be a feedback loop. I’ve found the best teams treat their category plan like a product roadmap—quarterly reviews, clear owners, measurable outcomes, and a prioritized backlog of initiatives.

Where AI fits (and where it doesn’t)

AI doesn’t replace segmentation. It makes segmentation less arbitrary.

• Helps: continuously updating risk signals (financial health, delivery performance, incident reports), spotting demand pattern shifts, and identifying supplier dependencies.
• Doesn’t: decide your risk appetite, trade-offs, or stakeholder priorities. That’s executive judgment.

If your CatMan “format” is a PDF no one opens after approval, AI will just make a nicer PDF faster. The win comes when AI feeds a monthly (or even weekly) category cockpit that drives decisions.

2) The best category strategies start with demand clarity, not supplier shortlists

Answer first: Demand management is the fastest way to improve category outcomes because it reduces noise before you negotiate.

Procurement often jumps straight to the supply side—RFPs, auctions, supplier rationalization—because it feels controllable. But many categories are “expensive” because demand is messy:

• Too many SKUs
• Uncontrolled specs
• Duplicate suppliers used by different sites
• Spend split across cost centers with conflicting requirements

When demand is unclear, you get predictable consequences: low compliance, frequent exceptions, and constant “urgent” buys that bypass strategy.

A practical demand-first workflow

For your next category refresh, run this sequence before you touch the supplier market:

1. Define the demand unit (per patient, per production batch, per employee, per shipment).
2. Normalize specs (what’s truly required vs. habit).
3. Separate recurring vs. project demand (they need different sourcing approaches).
4. Quantify variance across plants/regions (where standardization is feasible).

How AI improves demand visibility

AI can do real work here, especially in indirects and complex direct materials:

• Spend classification automation to reduce “miscellaneous” spend and improve category baselines.
• Forecasting and anomaly detection to flag sudden demand spikes (which often hide waste or policy drift).
• Specification clustering to identify near-duplicate items that prevent aggregation.

A useful stance: don’t aim for a perfect forecast—aim for faster detection of change. In volatile categories, early signals beat precise long-range predictions.

3) Stakeholder alignment is a category skill, not a soft add-on

Answer first: Category management fails more often due to stakeholder misalignment than due to weak negotiation.

You can build a sophisticated category strategy and still get crushed by a simple reality: the business doesn’t follow it. In regulated industries like biopharma (where leaders such as Sanofi’s procurement executives operate), stakeholder constraints are real—quality, validation, continuity, and compliance can outweigh price.

Category management best practice is being explicit about trade-offs:

• What are we optimizing—cost, resilience, innovation, speed, or sustainability?
• What are the non-negotiables (e.g., dual sourcing, audit rights, validated materials)?
• What decisions can be localized, and what must be global?

Make alignment measurable

Instead of “stakeholder buy-in,” use metrics procurement can actually manage:

• Contract compliance rate (by site, by business unit)
• Exception rate (and top three root causes)
• Cycle time from request to PO
• Savings realization vs. negotiated savings

AI’s role: less arguing, more evidence

AI can reduce political friction when it creates a shared fact base:

• Auto-generated should-cost ranges using input indices and historical purchase prices
• Supplier performance scorecards built from OTIF, defect rates, and incident history
• Risk heatmaps that show exposure by site and product line

Here’s what works: use AI outputs as inputs to stakeholder workshops, not as verdicts. People accept uncomfortable decisions faster when they can see the logic and assumptions.

4) Supplier strategy should be dynamic: segment by dependency, not just spend

Answer first: Modern supplier segmentation should reflect dependency and substitutability—because those drive risk more than spend size.

The classic “strategic vs. non-strategic supplier” language often hides the real question: What happens if this supplier fails next month?

Two suppliers can have the same annual spend, but very different risk profiles:

• Supplier A: easily replaceable, standard item, many substitutes.
• Supplier B: unique process capability, long qualification time, embedded IP.

Category management best practice is to build category-level resilience design:

• Single vs. dual sourcing decisions grounded in qualification lead time
• Inventory buffers where switching isn’t feasible
• Contractual levers (audit rights, capacity reservations, step-in clauses)
• Joint business plans for innovation where it matters

How AI supports supplier risk management

AI adds value when it connects fragmented signals:

• Early-warning indicators (shipment delays, financial stress patterns, quality drift)
• Network risk mapping (sub-tier dependencies, common logistics nodes)
• Scenario modeling (what if port X is disrupted; what if input Y spikes 18%)

A strong rule: If you can’t explain your segmentation in one sentence, it won’t drive behavior.

Example segmentation statement (one-liner):

“We prioritize suppliers by time-to-switch and qualification complexity, then tailor contracting, inventory, and SRM accordingly.”

5) Digital category management isn’t a dashboard—it’s decisions at the right cadence

Answer first: Digitizing category management means building repeatable decision cycles, not just implementing procurement analytics.

Many procurement orgs already have analytics tools, supplier portals, and spend cubes. Yet category decisions still happen via email threads, quarterly business reviews that drift, and urgent escalations. The missing piece is usually cadence and ownership.

A simple cadence model (that doesn’t overwhelm teams)

Pick the cadence that matches category volatility:

• High volatility / high risk categories: monthly category cockpit
• Stable categories: quarterly performance and demand review
• Long-cycle categories (capital, complex services): milestone-based steering

Each session should answer the same three questions:

1. What changed since last time? (demand, supply, risk, compliance)
2. What decisions are required now? (re-source, rebalance volumes, adjust specs)
3. What actions do we take and who owns them? (with dates)

AI can automate the boring parts—so humans do the hard parts

The best use of AI in procurement analytics is to reduce manual work:

• Automatically summarizing KPI movements and exceptions
• Generating supplier performance narratives from underlying metrics
• Flagging contracts at risk of leakage (price variance, off-contract buying)

But don’t skip governance. AI-generated insights without a decision forum become another notification stream everyone ignores.

A practical blueprint: make your next category plan “AI-ready” in 30 days

Answer first: You don’t need a full AI transformation to improve category management—start with data discipline, decision cadence, and two high-value AI use cases.

Here’s a realistic 30-day plan I’d recommend to a procurement leader who wants results before Q1 planning locks:

Week 1: Fix the baseline

• Choose 1–2 priority categories (high risk or high leakage)
• Clean the top 20 suppliers and top 50 items/services in the spend file
• Define 5 KPIs that matter (cost, service, risk, compliance, cycle time)

Week 2: Establish the decision forum

• Set the cadence (monthly or quarterly)
• Name a category owner and a stakeholder sponsor
• Define what counts as a decision vs. an update

Week 3: Pilot two AI use cases

Pick two that directly support decisions:

• AI spend classification + exception detection (to reduce leakage)
• Supplier risk monitoring + early-warning alerts (to prevent surprises)

Week 4: Turn insights into actions

• Create a prioritized backlog (top 5 initiatives)
• Assign owners, dates, and expected impact ranges
• Document assumptions and what would trigger a strategy change

This approach keeps category management grounded in business outcomes while letting AI carry the data-heavy load.

People also ask (and the straight answers)

Does the Kraljic matrix still apply in 2025?

Yes, as a teaching tool and a quick segmentation lens. No, as the final word. You need dynamic inputs (risk signals, demand variability, switching lead time) to keep it relevant.

What’s the best AI use case for category management?

Start with spend classification and leakage detection if your data is messy, or supplier risk monitoring if continuity is your biggest concern. Both create fast, visible wins.

How do I measure category management maturity?

Look at compliance, exception rates, decision cadence, and realized value—not the number of strategies published.

Where category management is heading in 2026

Category management is becoming a system of continuous choices: how you balance cost, resilience, compliance, and innovation category by category. AI will accelerate that shift, but only if procurement teams design decision rhythms that turn signals into action.

If you’re building your 2026 procurement roadmap now, start with one category where surprises are expensive. Make it measurable. Make it repeatable. Then scale what works.

What’s the one category in your org where a small improvement in demand clarity or supplier risk visibility would pay for the whole AI investment?