In November 2025, Gartner published a prediction that barely made a ripple outside the analyst community but should have set off alarms in every sales department on the planet: by 2028, AI agents will intermediate more than $15 trillion in B2B spending — with a growing share of those transactions involving no human decision-maker at all.
Let that number breathe for a moment. Fifteen trillion dollars. Routed through software that evaluates vendors, negotiates pricing, assesses compliance, and executes purchases autonomously.
This isn’t science fiction dressed up in a research note. It’s already starting.
The Shift From Assisted to Autonomous
Don Scheibenreif, a distinguished vice president analyst at Gartner, has been tracking the evolution of AI in enterprise procurement for three years. His team’s research charts a clear trajectory: from AI that helps humans make decisions, to AI that makes decisions and informs humans, to AI that makes decisions without humans needing to know.
“We’re moving from copilots to autopilots,” Scheibenreif said in a briefing accompanying the prediction. “The first wave of enterprise AI augmented human decisions. The next wave will replace them — starting with the decisions that are high-frequency, data-rich, and rules-based.”
Software procurement fits that description perfectly. Evaluating whether a SaaS tool meets security requirements, comparing pricing across vendors, checking license terms against corporate policies, running usage analytics to determine if a renewal is justified — these are exactly the kinds of structured, repeatable tasks that agentic AI handles better than humans.
Gartner’s broader data reinforces the acceleration. By 2026, 40% of enterprise applications will feature task-specific AI agents, up from less than 5% in 2025. By 2028, at least one-third of all enterprise software will incorporate agentic capabilities. And at least 15% of day-to-day work decisions will be made autonomously by AI — up from effectively 0% in 2024.
What This Means for How Software Gets Sold
If a quarter of enterprise software purchases are being made by AI agents within two years, the entire go-to-market playbook for B2B software needs to be rewritten.
Think about what a typical enterprise software sale looks like today: demo calls, POC environments, champion-building inside the buyer’s organization, executive sponsorship, procurement negotiations, legal review. The process can take months. It’s relationship-intensive, narrative-driven, and deeply human.
Now imagine the buyer is an algorithm.
AI agents don’t care about your brand story. They don’t respond to charisma on a demo call. They evaluate structured data — API documentation, security certifications, uptime SLAs, pricing transparency, integration compatibility. Vendors whose product information is clean, machine-readable, and verifiable will have an enormous advantage over those that rely on human persuasion.
Brent Adamson, a former Gartner distinguished advisor and co-author of The Challenger Sale, sees this as the biggest disruption to B2B sales since the internet.
“For 30 years, we’ve optimized sales processes around human psychology — building trust, creating urgency, managing stakeholder alignment,” Adamson told me. “None of that matters when the buyer is a piece of software evaluating your offering against 47 competitors in 12 seconds.”
The Procurement Transformation
On the buyer’s side, the change is equally profound.
Today, most enterprise procurement teams are overwhelmed. The average company manages hundreds of vendor relationships across dozens of software categories. Renewal cycles overlap. Usage data is scattered. Contract terms vary wildly. The result is a process that’s slow, inconsistent, and leaky — companies routinely overpay for underused software because nobody has time to audit every contract.
AI agents solve this by operating continuously. They monitor usage patterns in real time, flag underperforming tools, identify when contract terms drift out of alignment with market rates, and surface alternatives proactively. The context-switching cost that plagues human procurement teams — jumping between spreadsheets, vendor portals, and approval workflows — simply doesn’t apply to an AI that processes all of it simultaneously.
Gartner expects these agentic systems to rely on standardized trust frameworks — essentially, machine-readable credentials that allow AI agents to verify a vendor’s claims about security, compliance, and performance without human validation. If you’ve ever waited three weeks for a vendor security questionnaire to come back, you can see why this matters.
The 18-Month Window
Here’s the part that should keep managers awake. Gartner also predicted that by April 2026, most enterprises will begin abandoning assistive AI in favor of outcome-focused agentic workflows. The transition window is narrow.
Organizations that treat AI agents as a future concern rather than a present one will find themselves on the wrong side of a rapid shift. The companies building agentic procurement capabilities now — training their AI on internal purchasing data, establishing trust frameworks with vendors, defining the rules and boundaries within which agents can operate autonomously — will have a structural advantage that late movers will struggle to close.
And the stakes aren’t symmetrical. Gartner also noted that over 40% of agentic AI projects will be canceled by the end of 2027, most due to poor planning, unclear governance, or a failure to define where human oversight is actually needed. The winners won’t be the companies that adopt fastest. They’ll be the ones that adopt most thoughtfully.
“The question isn’t whether to deploy AI agents,” Scheibenreif said. “It’s whether you’ve done the governance work to deploy them responsibly. An autonomous purchasing agent with no guardrails is a liability. One with clear boundaries and good data is a competitive weapon.”
What Managers Need to Understand
For the average manager — someone running a team, managing a budget, making technology decisions — the practical implications break into three categories.
First, if you sell software, your product needs to be AI-readable. Structured data, transparent pricing, machine-verifiable claims. The era of winning deals through relationship selling isn’t ending tomorrow, but the percentage of deals decided by human judgment alone is about to shrink dramatically.
Second, if you buy software, start thinking about your procurement process as a data problem. The AI agents that will handle purchasing decisions need clean historical data — what you bought, what you paid, how much it was used, what worked and what didn’t. Companies that haven’t centralized this data will be feeding their agents garbage, and getting garbage decisions in return.
Third, and most importantly, define the boundaries now. Which purchasing decisions can an AI agent make autonomously? Which require human review? At what dollar threshold does a human need to sign off? These aren’t technical questions. They’re management questions — and the organizations that answer them before the technology forces the issue will be far better positioned than those that scramble to answer them after.
Gartner’s $15 trillion prediction isn’t about procurement. It’s about the first large-scale transfer of organizational decision-making from humans to machines. Software purchasing is just where it starts — because the data is structured, the stakes per transaction are manageable, and the rules are relatively clear.
What comes after procurement — hiring decisions, strategic investments, partnership evaluations — is a conversation most organizations aren’t ready to have. But the clock started the moment the first AI agent placed its first purchase order.
Eighteen months. That’s the window Gartner’s data implies. Not to adopt the technology. To decide how you want it to work before it’s working without you.