By 2028, Gartner predicts 90% of B2B buying will be intermediated by AI agents, channeling more than $15 trillion in spend through automated, machine-to-machine exchanges (Gartner, October 2025). It's one of ten strategic predictions Gartner presented at its IT Symposium/Xpo, and the most consequential one for anyone who sells to other businesses: procurement is shifting from a human-negotiated process to a software-negotiated one, at a speed and scale finance and sales teams have never had to plan for.
For sellers, the headline number matters less than what sits underneath it: if a buyer's software is making the purchase decision, your credit terms, payment flexibility, and settlement speed have to be readable and actionable by that software — not just by a human procurement manager.
Key takeaways
- Gartner expects AI agents to intermediate 90% of B2B purchasing by 2028, worth over $15 trillion in spend, replacing manual RFQs, negotiation, and purchase orders with machine-to-machine transactions (Gartner, 2025).
- Gartner separately forecasts that by 2030, 20% of monetary transactions will be programmable — carrying embedded terms and conditions so AI agents can transact with real economic agency (Gartner, 2025).
- Payments and credit terms are becoming a visibility factor, not just a back-office function: analysts note that AI purchasing agents prioritize suppliers with clear, machine-readable financing terms and predictable settlement over those relying on manual quotes and negotiation (PYMNTS Intelligence, February 2026).
What Gartner is actually predicting
Gartner's forecast is one of ten "strategic predictions" the firm issues annually for IT and business leaders, spanning talent, sovereignty, and what it calls "insidious AI." The B2B purchasing prediction sits alongside a few others directly relevant to commerce and payments:
- By 2028, organizations using multiagent AI for 80% of customer-facing processes will outperform competitors that don't (Gartner, 2025).
- By 2028, 90% of B2B buying will be AI agent intermediated, moving over $15 trillion in spend, with "verifiable operational data" becoming a form of currency and digital trust frameworks a prerequisite for participation (Gartner, 2025).
- By 2030, 20% of monetary transactions will be programmable, giving AI agents genuine economic agency to negotiate and settle deals autonomously (Gartner, 2025).
- By 2027, agentic AI reinvention will cut the cost-to-value gap in process-centric service contracts by at least 50% (Gartner, 2025).
Taken together, the throughline is that procurement, financing, and settlement are collapsing into a single automated decision, rather than three separate steps handled by three separate teams.
Why this changes the payments conversation, not just the sales conversation
Historically, B2B sellers competed on price, catalog, and relationship. Analysts covering the shift toward agentic commerce argue that's no longer sufficient: procurement agents evaluate suppliers on how well their pricing, financing terms, and fulfillment data can be parsed and trusted programmatically, and will route demand to a competitor before a human on the supplier's team even knows the opportunity existed (PYMNTS Intelligence, February 2026).
That reframes trade credit and BNPL from a checkout feature into a discovery signal. An AI purchasing agent comparing suppliers isn't going to wait on a multi-day credit application — it needs an instant, API-accessible answer on whether flexible terms are available and what they cost, at the moment it's evaluating options.
The trust and fraud problem nobody has fully solved
Faster, more autonomous transactions also mean faster, more autonomous fraud. Payments networks have already flagged rapidly escalating fraud exposure as agentic commerce scales, since the same automation that speeds up legitimate buying also speeds up bad actors probing for weaknesses in underwriting and identity verification (Digital Commerce 360, November 2025). Gartner's own prediction acknowledges this tension directly, noting that fragmented standards and security vulnerabilities in programmable money infrastructure will slow adoption even as the underlying technology matures (Gartner, 2025).
This is precisely why real-time credit and fraud decisioning — not just real-time payment processing — is becoming the harder, more important problem to solve. A seller that can approve a buyer (human or agent) in seconds, with fraud checks built into that same decision, is structurally better positioned for an agent-mediated market than one relying on manual review cycles.
How sellers should prepare
Three moves matter more than the rest, based on where analysts see the friction concentrating:
- Make your credit and payment terms machine-readable. If an AI agent can't parse your financing terms as cleanly as your price list, it will move on.
- Push credit decisioning to real time. Multi-day approval cycles are incompatible with a purchasing process that may complete in seconds.
- Treat fraud and credit risk as one decision, not two. As transaction speed increases, so does the value of instant, unified underwriting that a provider — not the seller — is accountable for.
This is the same infrastructure question B2B BNPL providers (Like Two) have already been solving for human checkout: instant approval, upfront seller payment, and provider-owned risk. Agentic commerce just raises the stakes and compresses the timeline.
This is the exact problem Two was built to solve, just for a new class of buyer. Two's Risk-as-a-Service engines already make real-time credit and fraud decisions at checkout — approving the large majority of business buyers in seconds, with Two (not the seller) taking on the non-payment risk. That combination — instant, API-accessible credit decisions plus fraud detection built into the same decisioning layer — is precisely the infrastructure an AI purchasing agent needs to transact with a seller instead of routing around it. As buying shifts from people to agents, sellers with financing terms that are fast, transparent, and machine-readable will be the ones agents can actually complete a transaction with.
FAQs
Is Gartner's $15 trillion figure the total size of B2B commerce, or just AI-agent-handled spend?
It's the portion of B2B spend Gartner expects to be intermediated by AI agents specifically by 2028 — not the entire B2B commerce market, which is far larger.
Does this mean human buyers disappear from B2B purchasing?
Not entirely. Gartner's related prediction is that multiagent AI will handle 80% of customer-facing processes by 2028, with humans focused on complex or high-stakes exceptions rather than routine transactions.
What does "programmable money" mean in this context?
Gartner defines it as monetary transactions that carry embedded terms and conditions, allowing AI agents to negotiate, transact, and settle payments with genuine economic agency — Gartner expects 20% of transactions to work this way by 2030.
Why does trade credit matter to an AI purchasing agent?
Because financing terms, not just price, are inputs an agent evaluates when comparing suppliers. Sellers offering instant, API-accessible credit decisions are easier for an agent to transact with than those requiring manual approval.
