AI procurement operating model

What happens when procurement activity grows 5× but your team does not?

Give buyers more capacity for repeatable execution while people retain strategy, relationships and judgment.

Thought experiment · not a forecast

1×

1×2×3×4×5×

People

Agents

Not part of the operating model

People’s workload vs. today’s capacity

Within capacity

Repeatable executionDecisions & relationships

Something falls behind, or you add people. Historically, those were the only two options.

01 · Why work grows

Growth multiplies the chasing, copying, checking and reconciling.

Every new site, acquisition and SKU adds more work for procurement. Tactical execution absorbs the capacity your team needs for sourcing strategy and supplier relationships.

  • New sites
  • Acquisitions
  • Suppliers
  • SKUs
  1. +More RFQs
  2. +More POs and checks
  3. +More supplier communications
  4. +More reconciliation
  5. +More fragmented data

The result

Execution crowds out strategy.

Less time for sourcing decisions, negotiation and supplier relationships.

02 · The hybrid operating model

People own the decisions. Agents work underneath them.

Select an agent to see what it does and where its work goes for approval.

VP / CPO

Priorities · ownership · governance

Directors & category leaders

Ownership · approvals · exceptions

Buyers & category teams

Strategy · relationships · negotiation · exceptions · approvals

▲ Approvals & exceptions go to people

RFQ agent

Does

Prepares RFQs from approved specs, routes them to approved vendors and normalizes the responses.

Flags and escalates to

Category buyer who owns the part

Waits for a person to

Approve a supplier or source change, or accept a recommendation outside guardrails.

Agents absorb

Quote · chase · check · reconcile · flag · escalate

People own

Strategy · relationships · negotiation · exceptions · approvals

03 · From processes to reusable skills

Approve a way of working once. Reuse it at every site.

  1. 01

    Data

    Parts, suppliers, contracts and spend.

  2. 02

    Codified skills

    Your approved playbook as steps, rules and exceptions.

  3. 03

    Agents

    Execute the skill consistently.

  4. 04

    Alerts · approvals · audit trail

    People stay in control and every action is recorded.

Example

Shared RFQ follow-up skill

Site A

Approved here
Owner
Site A category buyer
Approval queue
Site A category lead
Audit trail
Same format as Site A

Site B

Not yet

Not yet using the skill. Local rules and contacts are added before go-live.

Site C

Not yet

Not yet using the skill. Local rules and contacts are added before go-live.

Site D

Not yet

Not yet using the skill. Local rules and contacts are added before go-live.

Each site keeps a local owner and its own approval queue. The approved way of working, its controls and its audit format travel with the skill.

04 · Guardrails and accountability

Decide what agents may do alone, and what waits for a person.

  1. TIER 1

    Auto

    Agents act independently: gather data, send scheduled follow-ups, run defined checks, normalize information.

  2. TIER 2

    Review

    Agent prepares, a person reviews: supplier recommendations, discrepancies, alternative sources.

  3. TIER 3

    Approve

    A human decision is required: supplier changes, commercial decisions, pricing updates, policy exceptions.

Exception queue

Monitoring · example

  • Supplier response overdue on an open RFQ

    → Site 2 buyer

  • Price variance above the defined threshold

    → Category owner · fasteners

  • New supplier proposed as an alternative

    → Director approval

  • PO quantity doesn’t match the receipt

    → Site 3 buyer

Alerts route to the responsible category or site owner. Nothing in the queue resolves itself.

05 · How to prove capacity

Prove it on one workflow before estimating the rest.

  1. Step 1: Baseline

    Record buyer hours, volumes and cycle times today.

  2. Step 2: One workflow

    Run one skill with a clear owner and guardrails.

  3. Step 3: Measure

    Compare capacity and quality against the baseline.

  4. Step 4: Then expand

    Reuse the skill at other sites and workflows.

Capacity

  • Hours saved
  • Workflow throughput
  • Spend coverage
  • Service levels

Alongside quality

  • Output quality
  • Approval compliance

Capacity gained by skipping approvals isn’t capacity.

Where could your team create capacity first?

Step 1 · Plan

How to start deploying AI agents in procurement

Choose a workflow, map the data, set approval gates and measure the result before expanding.

Read the deployment guide

Step 2 · Optional

Choose one workflow to prove capacity

In a short working session, we look at which repetitive workflows grow fastest as your sites and spend expand, and pick the first one to measure.

Request a working session

More procurement work, without proportional headcount. People stay accountable.

Explore the hybrid team