Skip to content

Market shift

Agents already shape commercial evaluation. A person still signs.

Procurement agents already intake, shortlist and negotiate at machine speed. Most sellers still answer with PDFs, static rate cards and a deal desk in Slack, all built for people. Fully autonomous buying is still a thin slice of commerce, but the evaluation that decides the shortlist has already moved to agents.

See the product About PairRail

The mismatch

Why seller pricing breaks when agents ask

  • PDFs and rate cards get guessed at

    An agent has to scrape or parse them, and it fills gaps where the page is vague. Two agents reading the same page can come back with two different numbers.

  • Slack deal desks run at human speed

    An agent comparing ten sellers does not wait for a thread reply. A reply that lands tomorrow can arrive after the shortlist is made.

  • Rules written for people get routed around

    “Deals above $5,000 need a manager” stops a rep. Software can split the order, or ask again at slightly different sizes until the line shows.

  • Every protocol wants its own feed

    MCP, UCP, ACP, A2A and UAP each expect a different shape. Keeping a copy of the price book per protocol in step by hand is how prices drift.

Atlas answers all four from one governed catalog. See how it works or score your pricing.

Market layers

Where agents show up today

  • PairRail estimate

    Under 3%

    of completed purchase volume is fully autonomous purchases.

  • PairRail estimate

    20 to 30%

    of upstream discovery, pricing calculations, and commercial evaluations are directly driven or influenced by agents. Our planning range, not a single census. See the sources note.

That gap between agents shaping the decision and agents closing it is the window Atlas is built for. See the sources note.

LayerPairRail estimateWhat it means
Fully autonomous purchases< 3% (est.)Zero-click close is still rare. Most of it is narrow API top-ups, programmatic spend, and scheduled infrastructure. Agents evaluate far more than they buy alone.
AI-referred / mediated checkoutGrowing fastThe agent finds and configures; a human still signs. US retail ecommerce via general AI platforms is forecast at about $20.9B in 2026 (nearly 4× 2025), still a small share of total ecommerce.1
Upstream discovery & catalog queries~20-30% (est.)Pricing pages, docs, and API catalogs increasingly see machine traffic from coding and procurement agents. That is where scraper fatigue becomes a seller problem, and where protocol endpoints win.
Enterprise pilots & production agentsWidespread pilotsMost large organizations are piloting or deploying agents somewhere. Scaled multi-agent production is still uneven.26 Buyers are converging on shared agency: humans set budget and policy; agents evaluate and quote.

What changed, and where it goes next

  • What changed

    Protocol consolidation.

    Adapters are replacing screen scraping: MCP across developer tooling, Google’s UCP for commerce journeys, and ACP for agent checkout with payments partners.3

    Where it goes, for Atlas

    B2B buying becomes agent-intermediated.

    Gartner expects AI agents to intermediate about 90% of B2B buying by 2028, moving over $15 trillion through agent exchanges.5 Atlas: versioned catalogs and protocol endpoints agents query without scraping.

  • What changed

    Higher-intent mediated traffic.

    Adobe reported about 670% AI traffic growth to US retail sites on Cyber Monday 2025.4

    Where it goes, for Atlas

    Fail-closed seller hardening.

    As agent traffic rises, prompt injection and pricing arbitrage follow. Atlas: a pricing lab where the same order always gets the same price, approval thresholds agents never see, deal-desk escalation.

  • What changed

    Bots put load on pricing pages.

    Sellers feel load and rate-limit pressure from bots harvesting price tables.

    Where it goes, for Atlas

    Identity and signing for machine buyers.

    Telling a real enterprise credit line from an unverified crawler becomes table stakes. Atlas: scoped credentials, access inbox, per-key quote caps, audit trail.

  • What changed

    Buyer-side agents shipped first.

    Pactum9 negotiates long-tail supplier contracts; Zip10, Coupa, Zycus and Levelpath11 run agentic buying workflows. Those agents still land on human pricing pages.

    Where it goes, for Atlas

    Delegated micro-budgets.

    Buyers assign monthly allowances to procurement agents for seats and token top-ups. Atlas: seller-approved machine keys, published quote authority, per-key velocity limits.

  • What changed

    Shared agency, not blind auto-buy.

    Mandates, allowances and human deal desks are the durable pattern, which matches how Atlas fails closed outside policy.

    Where it goes, for Atlas

    Payment stays with your billing.

    Atlas: sends the approved quote to your billing system. Your processor charges the buyer, and you keep tax and fulfillment.

Both sides

Buyers send agents. Sellers answer with pages.

  • Demand side: buyer platforms already send agents

    Pactum9, Zip10, Coupa, Zycus and Levelpath11 automate how a company requests, compares and negotiates.

    They do not publish the seller’s approved price book.

  • Supply side: sellers are opening machine channels

    Stripe7 and Salesforce8 ship agent paths for catalogs, payments and ordering. API-native sellers have long sold to machines via keys and top-ups.

    Still thin: unsupervised close of complex B2B deals.

  • Atlas is the seller control plane those agents should query

    PairRail does not replace Pactum, Zip, Coupa, Zycus or Levelpath, and is not a source-to-pay suite.

    One approved catalog for every protocol.

  1. eMarketer, AI shopping tools gain traction… (2026): US ecommerce sales via general AI platforms about $20.9B in 2026, nearly 4× 2025, or about 1.5% of retail ecommerce. emarketer.com (opens in new tab)
  2. Enterprise deployment figures vary by survey. For example, the Fortune AIQ Advisory Board survey (July 2025) reports widespread active agentic projects among respondents, while noting that many pilots do not reach full implementation. Read “most enterprises” as piloting or adopting somewhere, not as universal scaled autonomy. Fortune AIQ whitepaper (PDF) (opens in new tab)
  3. Google, Universal Commerce Protocol and agentic commerce tooling (compatible with MCP, A2A, and AP2); industry coverage of ACP (OpenAI / Stripe) and MCP as the agent-to-tool layer. blog.google (opens in new tab)
  4. Adobe Digital Insights holiday shopping data (Nov 1 to Dec 31, 2025): generative AI referrals to US retail sites up about 693% for the season, and about 670% on Cyber Monday. Growth rates are period-specific. news.adobe.com (opens in new tab)
  5. Gartner, top strategic predictions for 2026 and beyond (Oct 2025): by 2028, about 90% of B2B buying is expected to be AI-agent intermediated, moving over $15 trillion through agent exchanges. Intermediation is not the same as unsupervised purchasing. gartner.com (opens in new tab)
  6. McKinsey Global Survey on the state of AI: about 23% of organizations report scaling an agentic AI system somewhere in the enterprise, and about 39% say they have begun experimenting. Most scaling remains limited to one or two functions. mckinsey.com (opens in new tab)
  7. Stripe, Machine Payments Protocol and Agentic Commerce Suite: open payment coordination for agents, plus catalog and checkout paths for selling through AI agents. Public examples include Browserbase metering agent browser sessions. stripe.com (opens in new tab) · Agentic Commerce Suite (opens in new tab)
  8. Salesforce, B2B Commerce innovations (Agentforce Buyer Agent, GA June 2026): account-aware product discovery and ordering over channels such as WhatsApp and SMS. Salesforce cites Siemens running Agentforce B2B Commerce in over 25 countries. salesforce.com (opens in new tab)
  9. Pactum, autonomous supplier negotiation (public customer programs include Walmart and other large enterprises): multi-stage agents that renegotiate long-tail commercial terms with suppliers. Named results on Pactum’s site are vendor-reported. pactum.com/clients (opens in new tab)
  10. Zip, Introducing agentic procurement orchestration (2025): intake-to-pay orchestration in front of existing ERPs, plus a published suite of task-specific procurement agents. zip.com (opens in new tab)
  11. Source-to-pay and AI-native procurement suites shipping agentic intake, sourcing, and supplier workflows. Coupa Navi / Compose: coupa.com (opens in new tab). Zycus Merlin Agentic Platform: zycus.com (opens in new tab). Levelpath (AI-native procurement): levelpath.com (opens in new tab). These are buyer-side products. They illustrate demand hitting suppliers; they are not PairRail alternatives.

Notes

  • Sellers that stay human-HTML-only risk losing the shortlist when agents prefer machine-readable catalogs.
  • AI-driven referral to retail sites surged in recent holiday windows.4
  • The idea that agents would buy enterprise software without fences has faded. Buyers want agents to evaluate and people to commit.
  • Procurement AI is not a forecast. It is a crowded buyer category. Pactum runs multi-stage negotiation agents against long-tail supplier contracts.9 Zip sits in front of the ERP as intake and agentic orchestration.10 Coupa and Zycus add agentic intake, sourcing, and spend analytics inside source-to-pay; Levelpath embeds agents in contract discovery and supplier talks.11
  • Stripe is shipping agent payment and catalog paths so services can sell to agents (for example Browserbase metering sessions).7 Salesforce’s Agentforce B2B Buyer Agent lets account-aware ordering happen over WhatsApp and SMS; Siemens is cited as running Agentforce B2B Commerce across many countries.8
  • Buyer suites own intake, policy, and negotiation for the company that is spending. Sellers still answer with a pricing page, a PDF, or a deal desk. When a procurement agent shortlists you, it should read your catalog, not scrape your site.

The under-3% autonomous-purchase / about 20-30% discovery framing is PairRail’s operating synthesis for seller planning, not a single audited market census. Analyst forecasts (including Gartner’s B2B intermediation outlook) describe direction and timing, not Atlas revenue. Named buyer platforms (Pactum, Zip, Coupa, Zycus, Levelpath) and seller rails (Stripe, Salesforce) illustrate that machine-facing commerce is shipping. They are not PairRail partners, customers, or competitors unless stated elsewhere. PairRail’s job is the seller catalog those buyer agents query. Use this page to prioritize catalog and protocol readiness before chasing fully autonomous checkout.

Score a commercial sheet.

See how a public price page reads to the same scorer Atlas uses.

Score your pricing, free