By 2028 You'll Have 10 AI Agents for Every Seller. Most Won't Help.
Gartner predicts AI agents will outnumber human sellers 10 to 1 by 2028, and that fewer than 40% of sellers will say those agents actually improved their productivity. The gap between agent count and agent value is the whole story.
Every sales org I talk to is either deploying AI agents or being told they should be. Almost none of them are asking the question Gartner's own research says actually matters: whether more agents translates into more output at all.
Gartner predicts that by 2028, AI agents will outnumber human sellers 10 to 1 inside sales organizations. In the same forecast, it predicts fewer than 40% of sellers will say those agents actually improved their productivity. Read those two numbers together and the finding is almost uncomfortable in how directly it contradicts the pitch every AI sales tool is currently making: agent count and agent value are not the same curve, and most organizations are about to learn that the hard way, after they've already bought the agents.
Dan Gottlieb, the Gartner VP Analyst behind the research, put the mechanism plainly: "more agents will not automatically mean more productivity." Without the right data foundation, workflow integration, and seller experience, the result is what he calls agent sprawl, more digital activity happening somewhere in the stack, with little improvement in what a seller actually gets done. Gartner's own upside case is specific too: sales leaders who overhaul their data and workflow foundations first are predicted to be five times more likely to see real ROI from AI than those chasing quick fixes.
Why more agents doesn't mean more output
This isn't really a new lesson, it's an old one wearing an AI label. Adding headcount to a broken process has always produced more activity without more results, more calls into a bad list, more emails through a leaky funnel, more meetings that don't move anything forward. An AI agent is headcount that's nearly free to add, which means the temptation to add ten of them instead of fixing the process underneath is stronger than it's ever been with human hiring, where cost at least forced some discipline. Gartner's forecast is describing what happens when that discipline disappears: agent sprawl, not agent leverage.
The tell is in what the 60% of sellers who won't report a productivity gain are actually experiencing. It's not that the agents don't work in isolation, an agent that qualifies leads, drafts a follow-up, or summarizes a call transcript genuinely does that task. It's that the task being automated was never the actual bottleneck. If the real constraint is that reps don't know what to say when a real objection lands, or don't ask the right question on discovery, an agent that speeds up scheduling or note-taking around that moment doesn't touch the thing that was actually costing deals.
What the 40% are probably doing differently
Gartner's own recommendation points at the answer: fix the foundation, the data an agent works from, how it's actually integrated into a rep's real workflow, and the experience a seller has using it, before adding more agents on top. Practically, that means being specific about which task is genuinely a bottleneck before automating it, rather than automating whatever's easiest to automate first. Scheduling and note-taking are easy to automate and rarely the constraint. The moment a rep is live with a skeptical prospect, deciding what to say next, is usually the actual constraint, and it's also the one task no agent can currently do for a human seller in real time.
That's not an argument against AI agents. It's an argument for being honest about which 40% you're trying to join. Live practice on the actual conversation, the part still entirely on the human seller regardless of how many agents sit around it, is where the productivity gain Gartner's data says most sellers won't get actually has to come from.