RESHUFFLE An interactive companion to the book

Procurement Observatory

How agentic AI transforms procurement

What this Observatory examines

Agents are not only automating procurement tasks. They are reorganizing the buying process, changing how firms make decisions, and opening procurement to new competitors.

The Observatory examines this change at three levels: the workflow, the firm and the competitive ecosystem.

The reframe

AI changes procurement at three levels

Workflow

From automating tasks to reorganizing the buying process

Most AI programmes improve individual steps. Agents can coordinate requirements, supplier options, rules, evaluation and execution across the buying journey.

  • Separate tasks
  • Coordinated journey

Firm

From final approval to the architecture of the decision

A person may still approve the purchase while agents and providers shape the options, apply the rules and execute what follows.

  • Shape
  • Apply
  • Approve
  • Execute
  • Record

Approval is one position within a larger decision architecture.

Competition

From procurement tools to competitors entering from the side

Payments platforms, marketplaces, contract systems and systems of record can enter procurement from capabilities they already control.

  • Adjacent position
  • Procurement control point

What the evidence shows

These changes are already becoming visible

Workflow

Cross-stage coordination is already visible

What we examined

Agentic procurement offerings across the buying journey, and, separately, the supported interactions between them.

Breadth analysis

  • 54agentic procurement offerings examined
  • 486offering-stage observations
  • 184demonstrated effects

Coordination analysis

  • 17cases of workflow coordination

The important signal is not simply that agents appear throughout procurement. Some are beginning to connect work that previously required manual handoffs.

Examine the workflow evidence

Firm

Approval remains with the firm, but much of the decision is configured elsewhere

What we examined

49 decision positions across 10 procurement areas.

Who configures the position

  • 5Firm
  • 34Providers
  • 7Institutions
  • 3Counterparties

All 5 firm-configured positions occur at approval.

The firm may retain the formal decision while depending on others for the capabilities that prepare, constrain and execute it.

Examine the firm evidence

Competition

New competitors are entering from positions they already control

What we examined

15 candidates that could enter procurement from capabilities they already hold.

Three separate units

  • 15candidates examined
  • 9actors represented
  • 12documented boundary-crossing events

The emerging contest is not only between procurement applications. Adjacent platforms are extending from payments, supplier access, contracts, data and records.

Examine the competition evidence

The executive decision

What should your company do differently?

Workflow

Redesign the operating model around coordination

Where are people still managing handoffs that an agent could coordinate across the buying process?

Design your AI-native operating model

Firm

Decide which capabilities your firm must retain

Which decision capabilities should you own, configure with a provider or safely delegate?

Design your AI value stack

Competition

Map where new competitors can enter

Which platforms already control supplier access, rules, execution or records, and where could they move next?

Map your competitive ecosystem

Workflow redesign changes the capabilities the firm needs. Those capabilities shape its competitive position. New competitors can then force another redesign.

Evidence, definitions and methodology denominators, separations and what each count is not

Each figure on this page is a count of what was examined, not a market estimate. The 486 offering-stage observations and the 21 supported interactions are separate populations with separate denominators, and are never combined. The 49 decision positions are coded positions in 10 examined procurement areas, not a prevalence or an adoption rate. The 15 candidates, 9 actors and 12 events are three different units and are never divided into one another.

Cases of workflow coordination. The coordination analysis examined 21 supported interactions. Of those, 17 connect more than one part of the buying process, and 14 select a next action within stated limits. The 17 that connect more than one part are what this Overview calls cases of workflow coordination. They are a separate population from the 486 offering-stage observations and are not a subset of them.

How the evidence was produced