$500B
Walmart's annual spend when Pactum's first business case began
Kaspar / 02:12Field Notes / Interview series
AI Negotiation
Field Notes
A conversation with Kaspar Korjus about Pactum's digital workforce, AI negotiators in procurement, agent identity and the changing boundaries between human and machine work.
// 01
One Big Idea
“You can now change the company behaviour and direction by just chatting with an agent.”
Pactum's agents are moving beyond task automation into a digital workforce: negotiating with suppliers, checking requisitions, enforcing policy and translating a changed strategy into new behaviour across the enterprise.
// 02
By the Numbers
$100B
of that spend had no people available to manage it
Kaspar / 02:121% = $1B
the annual saving from finding one percent of improvement in that long tail
Kaspar / 02:1910–15 min
for an agent negotiation, compared with weeks of human back-and-forth
Kaspar / 04:495 days → 30 sec
a requisition review moving from a human SLA to an agent response
Kaspar / 05:021M
requisition checks completed since March
Kaspar / 05:314,000×
faster than manual reviews, according to Pactum's release
Kaspar / 05:3120
key projects in Estonia's AI initiative
Kaspar / 16:0275%
of participants in the 2022 pilot preferred negotiating with AI
Katja / 17:21 · historical2030
Kaspar's prediction for agents in board meetings as equal partners
Kaspar / 11:37200
policies a machine can check on every requisition
Kaspar / 14:2010,000
suppliers an agent can negotiate with at scale
Kaspar / 14:50// 03
The Brief
The conversation, separated into the ideas that matter for people building, governing and managing a machine workforce.
EvidenceProcurementExec
Procurement is already a digital workforce
Pactum's agents are doing the work of a procurement function: requisitions, sourcing, supplier outreach, negotiation and system updates.
Speed & scaleOperationsProcurement
Weeks become minutes
Negotiations that took weeks now take ten to fifteen minutes; a five-day requisition SLA becomes a thirty-second response.
Speed & scaleDataAutomation
Scale is the reliability advantage
A machine can check two hundred policies on every requisition and negotiate with ten thousand suppliers while humans prioritise the few they can reach.
Operating modelPeopleChange
The hard part is change management
The technology is capable. The difficult work is deciding which roles remain human, where agents work alongside people and how the transition happens.
Identity & governanceLeadershipRisk
Manage agents like employees
Give each agent a role, boundaries, measures of success and the data access it needs. A free workforce without context can harm a company rapidly.
Identity & governanceGovernmentTrust
Agents need their own identity
Estonia is exploring digital identities that let agents act with delegated rights without handing them a person's full access to banks, systems or government.
Negotiation & valuesCommerceValues
Price is only one term in the deal
Agents can exchange what each side values: payment terms, contract length, allowances, responsible sourcing and verifiable supplier credentials.
AccountabilityRiskBoard
The company is still accountable
An autonomous agent does not become a liability shield. The company or human whose authority it uses remains responsible for the action and its consequences.