I've now read the same slide three times this quarter, in three different vendor pitches: by 2028, AI agents will handle 90% of B2B buying. Every time, the pitch that follows is the same, spend engineering time now making your site agent-readable or get shut out. So I went and checked how many of my own prospects are actually being pre-qualified by an agent today. The honest answer was almost none. Here's the gap between that prediction and what's actually happening in 2026, and what I'm spending time on instead of chasing the headline.
The stat everyone is quoting
The number comes from Gartner's prediction that AI agents will intermediate more than $15 trillion in B2B purchasing by 2028, with 90 percent of B2B transactions involving an agent somewhere in the process within three years. It's a real forecast from a firm that gets taken seriously, and I'm not arguing the trendline is wrong. I'm arguing the timeline is being flattened in every pitch deck that cites it, from a three-year runway into an urgent now.
What adoption actually looks like right now
Industry surveys on agentic AI adoption inside the enterprises that are supposed to be doing this buying tell a different story than the 2028 headline. Roughly 30% of organizations are still just exploring agentic options, another 38% are piloting something, and only around 14% have a solution ready to deploy. The share actually running agents in production, the ones that would be pre-qualifying your vendor before a human sees your site, comes in closer to 11%.
That's not a rounding error next to 90%. That's the difference between a prediction for 2028 and the reality of 2026. Somewhere between piloting and production, a lot of these initiatives are also stalling out. Enterprise buyers aren't skeptical of what agentic AI could do, they're skeptical because they've already been burned. They've funded pilots that went nowhere, paid for tools nobody on the team adopted, and watched integration problems kill a promising rollout. Trust in AI outputs is flattening even as usage climbs, which is a strange combination if you assume adoption and trust move together.
Why the gap gets hidden in every pitch
Every vendor selling agent-readiness tooling has a reason to collapse three years into an emergency. Urgency sells retainers. But the founders I talk to who've actually gone looking, pulling raw server logs instead of trusting a GA4 dashboard, find real agent traffic on their pricing pages. It's just thin. A handful of crawler hits a week, not a flood of pre-qualified buyers replacing their sales funnel. The technology is real and it is growing. It is not yet the dominant buying channel for a mid-market or early-stage B2B SaaS company, whatever the 2028 slide implies about this quarter.
What I'm actually doing about it this year
I'm not ignoring the trend, and I'm not telling anyone to skip it. I'm just sizing the investment to match 2026 adoption, not the 2028 headline, which changes the list a lot.
- Do the cheap, permanent groundwork now: un-gate your pricing, add Schema.org Offer markup, rebuild image-based comparison tables as real HTML. None of this hurts human buyers, and it costs an afternoon, not a roadmap quarter.
- Check your own logs before believing anyone's adoption number, including mine. Agent traffic doesn't show up in GA4, you have to pull it from the CDN or server layer directly.
- Don't build a dedicated agent-facing microsite, a custom API for agent queries, or a full content strategy around agent optimization yet. That's roadmap-quarter money chasing an 11% audience, not a 90% one.
- Re-run the 3-question readiness test once a quarter instead of once. Adoption curves aren't linear, and the point where this flips from cheap-insurance to must-have will arrive faster than the 2028 date suggests once enterprise buyers stop treating pilots as disposable.
The framework, without the hype
Treat the Gartner number as a compass, not a deadline. It tells you the direction is real and the destination is big. It does not tell you that this quarter's roadmap needs an agent-first rebuild. The founders getting burned right now aren't the ones ignoring agentic buying, they're the ones who read a three-year forecast as a this-year mandate and diverted a headcount's worth of engineering time toward an audience that's currently 11% of the market they're selling into.
Frequently asked questions
Is the Gartner 90% by 2028 prediction credible?
It's a serious forecast from a firm whose predictions carry weight, and the underlying trend of AI agents participating in B2B research is real and growing. The issue isn't credibility, it's that a three-year forecast is routinely misquoted as describing this year's buyer behavior.
How much AI agent adoption is real today?
Survey data on enterprise agentic AI puts roughly 14% of organizations at deployment-ready and around 11% actually running agents in production, with the rest still exploring or piloting. That's the closest current proxy for how much real B2B buying already runs through an agent.
Should a small B2B SaaS company invest in agent-readiness at all?
Yes, but at the cost of an afternoon, not a quarter. Un-gate your pricing, add machine-readable schema, and check your own server logs for agent traffic. Skip anything that requires a dedicated build or a new headcount until your own data shows agent traffic is a meaningful share of your funnel.
How do I know when it's time to invest more?
Watch your own raw traffic logs, not industry-wide predictions. When agent hits on your pricing and docs pages become a consistent, growing share of pre-sale research traffic, that's your signal to invest further, whatever year Gartner's forecast is aiming at.
The forecast and the pitch deck built on top of it are different things. One is a reasonable three-year bet. The other is someone else's urgency, dressed up as yours.