Agent Sprawl: AI Sellers Are Multiplying Faster Than the Systems That Run Them

Every major sales platform now has an agent roadmap, and the agents are no longer just summarizing calls. They research accounts, write outreach, prep reps for meetings and draft proposals. For revenue leaders, the pitch is obvious: more capacity without more headcount. The research arriving this fall suggests a less comfortable outcome is just as likely. Agents are multiplying much faster than the data, workflows and oversight they need to be useful, and adding more of them to a weak foundation produces activity, not revenue.

Ten Agents Per Seller, and Little to Show

Gartner's latest prediction for sales organizations is blunt. By 2028, AI agents will outnumber sellers by ten to one, yet fewer than 40% of sellers will say those agents improved their productivity. Dan Gottlieb, a vice president analyst in the Gartner Sales practice, warned that chief sales officers "risk creating agent sprawl, with more digital activity, but little improvement in seller impact." Agents, he said, "are only as effective as the systems they operate within."

The research behind the prediction, a survey of 210 CSOs and senior sales executives fielded in January and February 2026, carries a second number that explains the appetite for agents. Sixty percent of CSOs said their revenue number is largely driven by factors outside their control. When the number feels out of reach, a tool promising more pipeline from the same team is hard to turn down. Gartner's counterweight is its prediction that CSOs who overhaul data, automation and user experience will be five times more likely to gain ROI from AI than those who choose quick fixes.

The supply side is accelerating regardless. On September 11, Salesforce introduced a slate of named Agentforce agents, including an outbound sales agent called Hunter that, according to SiliconANGLE, finds leads and drafts outreach emails and then supports later deal stages by preparing reps for presentations and creating proposals. Hunter is the first agent to run on what Salesforce calls a long-horizon runtime, which builds work plans and reuses data across sessions so the agent can stay on a deal for weeks. It launched alongside two other sales agents, Piper and Carter.

An agent that works a deal for weeks will act on whatever the CRM tells it. If the record is wrong, the agent is wrong at scale.

Volume Is Not the Same as Pipeline

Early evidence from sales development shows what happens when AI is pointed at volume. Orum's 2026 State of Sales Development, based on a survey of 300 sales professionals, found that most teams use AI for research, enrichment and list building, while very few use it to improve the conversations that actually create pipeline. Only 12% use AI for coaching. The report also found that the surge in automated outreach has reduced the effectiveness of traditional digital channels, and 80% of respondents said customer acquisition cost was up or unchanged from the prior year.

The organizations getting results are using AI differently. A separate Gartner survey of 227 CSOs, fielded in August and September 2025, found that sales organizations giving sellers AI-enabled next best actions are 2.6 times more likely to achieve commercial growth, and those prioritizing AI upskilling are 2.4 times more likely to achieve strong revenue growth. Greg Hessong, a senior director analyst at Gartner, said the most effective organizations "are not simply layering AI onto existing ways of working."

Buyers explain why the human half of the equation still matters. In the same release, Gartner's survey of 645 B2B buyers found that, compared with generative AI, buyers were 28 percentage points more likely to say a sales rep helped them advance to the next purchase step and 39 points more likely to say a rep understood their needs. Buying groups with low dysfunction were 13 times more likely to report high-quality deals. Agents can clear the path to that conversation. They have not shown they can replace it.

Nobody Is Watching the Agents

The governance gap is the part revenue leaders are least prepared for. A report from the nonprofit Partnership on AI, covered by The Christian Science Monitor on October 7, tested four agent systems and found six areas of incomplete oversight. Systems identify agents only by a name or number, and subagents inherit their parent's identity. Records were missing for human interventions, changes to an agent's permission mode and changes to its memory. Even complete telemetry shows what an agent did, not why it did it.

Madhulika Srikumar, head of AI safety at Partnership on AI, framed the problem in terms any sales leader will recognize: "Think of a company hiring 1,000 new employees." Then: "You haven't interviewed these people." In a revenue organization those unvetted hires send email under the company's name, reference pricing and draft proposals. Without an audit trail, a CRO cannot tell a buyer, a legal team or a board which agent said what, or on whose authority.

The CRO Playbook for Agent Sprawl

• Count your agents before adding more. Inventory every agent and AI feature touching pipeline, who owns it, what it can send and what data it reads. Sprawl starts when nobody has the full list.
• Fix the record agents act on. Agents inherit every gap in the CRM. Prioritize clean opportunity data, stage definitions and loss reasons before giving an agent a deal to work for weeks.
• Point AI at decisions, not volume. Gartner's growth link is to next best actions, not more outreach. Aim agents at prioritization and deal guidance before scaling sends.
• Govern what agents produce. Proposals, order forms and outreach should come from approved templates and pricing rules, with human approval on anything a buyer will sign.
• Demand an audit trail. Require vendors to log agent identity, permission changes and human overrides. If you cannot reconstruct what an agent did, you cannot defend it.
• Measure capacity, not clicks. Judge agents on seller capacity, win rates and deal quality, not on emails sent or hours saved.

The ten-to-one future Gartner describes is close to arriving whether revenue leaders plan for it or not, because the agents are shipping inside platforms teams already pay for. The question for CROs is whether those agents land on a foundation that turns their activity into revenue, or on one that simply generates more noise for buyers who are already tuning it out.