How AI Agents Are Reshaping B2B Procurement and Sales

Autonomous AI agents are moving beyond research support and into active roles in enterprise procurement: building shortlists, negotiating terms and, in advanced pilots, finalizing deals without human intervention. This shift strains commercial playbooks that were designed around human-to-human relationships, per-seat licensing and quarterly social renewals.

Concrete examples and measurements

• In 2021 Maersk, the world’s largest container shipping company, deployed agents from the startup Pactum to negotiate freight-lane contracts. The system operated fully autonomously—contacting carriers, running multiple negotiation rounds on price, route obligations and payment terms, and closing deals. It achieved a 96% agreement rate among carriers, and in controlled trials the agent secured rates 22% lower on identical lanes than human negotiators.
• Maersk reports $54 billion in annual revenue and the agents worked across more than 50 carrier partnerships; negotiations ran according to predefined parameters rather than human rapport-building.

Market trends and survey data (2025–2026)

• McKinsey (November 2025): a global survey of 1,993 enterprise respondents found that 62% are at least experimenting with AI agents and 10% have fully scaled agent capabilities.
• Cloudflare (July 2025): overall AI bot crawling grew 24% year-over-year, with agent-driven requests (automated traffic acting on behalf of users) the fastest-growing subcategory. Cloudflare’s CEO expects automated software traffic to overtake human-generated traffic by 2027.
• Gartner (August 2025): projects that by the end of 2026, 40% of enterprise applications will incorporate task-specific AI agents (up from <5% in 2025).
• G2 (April 2026): in a survey of more than 1,000 B2B software buyers, AI chatbots now top the list of sources that influence vendor shortlists; 51% of buyers start their research with AI chatbots, up from 29% the prior year.
• Forrester (2026): forecasts that at least 20% of B2B sellers will face AI-powered buyer agents this year.

The problem with per-seat licensing in a continuous-compute era

Per-seat pricing assumes discrete human sessions. When a delegated procurement agent runs workflows it spawns parallel sub-processes, operates continuously across time zones and executes at machine speed; seat counts no longer map cleanly to usage.

• Kearney estimates that AI procurement agents could erode up to 500 basis points of EBIT for distributors by commoditizing supplier selection and compressing average selling prices by approximately 8%.

Some sellers are already experimenting with outcome-based models:

• Sierra (backed by a16z) moved away from seat- or session-based pricing toward billing for measurable outcomes.
• Intercom’s Fin AI charges $0.99 per successfully resolved conversation while providing unresolved interactions free of charge; Archana Agrawal, President of Intercom, explained that customers did not want to pay for activity, so Intercom charges when a positive outcome occurs (GTMnow).

McKinsey’s February 2026 analysis found a chemicals company that deployed agents for autonomous sourcing of consumables achieved 20–30% efficiency improvements for procurement staff and a 1–3% increase in value capture.

Relationship-driven sales face new limits

Traditional negotiation assumes a human counterpart with memory, reputational stakes and susceptibility to persuasion. Agents, however, screen on measurable attributes: price, contract terms, delivery specs and compliance. If those things are not programmatically available, vendors risk being excluded from shortlists before a human relationship begins.

• SUEZ UK, part of the €19 billion SUEZ Group, used Pactum’s agents to reach 2,000 additional suppliers in two months, realizing average potential savings of 2.5% and cost reductions of 15% through competitive purchasing pressure.

Customer success is also becoming empirically driven. Agents calculate ROI from API usage logs, discover competitor pricing programmatically and recommend actions to human principals. Clari Labs analyzed 10 million opportunities across 121 major global enterprises (Jan 2023–Dec 2024) and found average contract value fell 50% year-over-year while average expansion deal cycles grew from 92 days to 125 days.

Forrester’s Buyers’ Journey Survey reports 68% of B2B buyers already have a front-runner vendor in mind at the start of a purchase process. In the age of AI agents, that research often happens in the background; competitors using Generative Engine Optimization (GEO) can become algorithmic front-runners before a Customer Success Manager even recognizes risk.

Sean Neville of Catena Labs notes in a16z’s 2026 trend report that in financial services, non-human identities already outnumber human employees 96 to 1. Entro Labs’ H1 2025 NHI Management report states non-human identities exceed human ones 144 to 1 across enterprises. These identities do not respond to relationship motions account management was designed for.

Practical steps for an agent-first go-to-market

B2B sellers should consider three critical shifts:

1. Adopt outcome-based pricing architectures

   • Transition from strict per-seat billing to consumption- or outcome-based hybrids so agents can evaluate providers on quantifiable results. McKinsey’s Feb 2026 agentic procurement analysis showed a telco reduced negotiation analysis and email time by up to 90% and achieved 10–15% savings across vendors with AI-guided negotiations.

2. Publish machine-readable product surfaces

   • G2 (April 2026) found 85% of B2B buyers rate a vendor more highly when an AI answer engine includes them. Universal Commerce Protocol (UCP) offers a way to expose product information, pricing, availability, terms and checkout flows in a format readable by AI systems. Google’s Agent2Agent and Agent Cards (structured JSON capability documents) are another mechanism; vendors without machine-readable capability profiles will increasingly become invisible to agent-driven shortlisting.

3. Implement agent-compatible authorization and verification

   • An AI agent cannot finalize transactions or commit budget without verifiable authority. Vendors must redesign checkout flows to programmatically verify an agent’s spending limits and legal liability. Venture funding and startups like Descope (expanded by Lightspeed Venture Partners) target the “agentic identity” problem to secure agent authentication and authorization.

Risks and contradictions

Deploying more AI does not automatically increase seller productivity. Gartner (November 2025) projects AI agents will outnumber human sellers tenfold by 2028, yet fewer than 40% of sellers report that AI improved their productivity. VP Analyst Melissa Hilbert cautions that piling more AI tools onto complex workflows can overwhelm sellers and accelerate burnout.

B2B sellers that treat agent buyers as a transient trend risk losing screening and sourcing decisions to counterparties they cannot engage. Instead of resisting agents, sellers must adapt their commercial architecture—pricing, product data and authorization—to remain visible and competitive in markets increasingly governed by AI-assisted procurement.