Gartner predicts AI agents will outnumber sellers 10 to 1 by 2028. What should sales operations do?

What Gartner predicted, and what kind of statement it is

Gartner's announcement of 28 July 2026 makes two predictions in one headline. AI agents will outnumber sellers 10 to 1 by 2028, and fewer than 40% of sellers will say agents improved their productivity.¹ Both are forecasts, published this week, about a year that has not happened yet.

The pairing is the interesting part. Gartner is not predicting a sales floor run by machines that works better. It is predicting a lot more machines and a sales force that, in most cases, will not feel the benefit. That is a warning about measurement as much as about technology.

Our view: the second half of the headline matters more to a sales operations lead than the first. Agent counts will rise whatever you do, because every CRM, dialler and email tool is adding them. Whether your sellers end up in the minority that feels more productive depends on choices you can make now.

Why would sellers not feel more productive?

Because most agents in sales will be added without a measure of what they replace. An agent that drafts follow-up emails, enriches records or books meetings produces activity. Activity is easy to count and easy to mistake for productivity.

Gartner's own explanation is agent sprawl: without the right data foundation, workflow integration and seller experience, more agents bring more digital activity but little improvement in seller impact.¹ From the way sales agents are usually rolled out, we would add three practical causes. Agents arrive bundled in tools rather than chosen for a process. Nobody records how long the work took before. And the seller still has to check the agent's output, so the time saved is partly spent reviewing.

A seller asked "did agents make you more productive?" answers from experience: more notifications, more drafts to approve, the same quota. Unless sales operations can show the seller a number, such as leads followed up within the hour or meetings booked per hundred leads, the honest answer is often "not really".

What should a lead follow-up agent log?

Five fields per lead, written to the CRM record, not to a separate dashboard. Lead follow-up is a sensible first sales agent because the volume is steady and speed is measurable. It also produces a clean before-and-after if you capture the right fields from the first day.

Our recommended log for a lead follow-up agent. Field names are ours; map them to your CRM.

With those five fields you can answer the questions a sales director will ask in month three. How much faster are leads touched? Of the replies, how many came from messages a seller edited? What does a booked meeting cost in agent spend? The same discipline applies to any first agent, and our scoring model for choosing the first process explains why volume and reversibility come first.

Keep the agent's permissions narrow while you collect this data. A follow-up agent needs to read lead records and create drafts or tasks; it rarely needs to change deal stages or delete contacts. Our least-privilege checklist for ERP and CRM agents covers the role design.

Where each log field is written

1. Lead arrives: trigger
2. Agent drafts: action, cost
3. Seller reviews: touch
4. Buyer responds: outcome
5. CRM record

What the adoption data says about readiness

Large companies are deploying agents faster than they are learning to see what agents cost. In KPMG's AI Quarterly Pulse, a survey of 204 US leaders at firms with $1bn or more in revenue (28 April to 25 May 2026), 53% said they were deploying AI agents, against 55% the quarter before.² In the same survey, 18% orchestrated multiple agents across workflows, up from 9%.²

Only 26% of those leaders reported full, real-time visibility of AI operating costs.² That is the cost field in the table above, missing at roughly three in four of the largest US firms. Our note on the KPMG pulse sets out the running-cost check to do before a second agent.

Across all organisations the picture is earlier still. Gartner's 2026 CIO and Technology Executive Survey (published April 2026; data period not stated) found that 17% of organisations had deployed AI agents.³ A 10-to-1 ratio by 2028 implies very fast growth from that base, most of it likely to arrive inside tools you already pay for.

What sales operations should do this quarter

Take a baseline now, before the next agent arrives in a tool update. Pull three months of lead data: time from arrival to first touch, reply rate and meetings booked per hundred leads. Without that baseline, no agent can ever be shown to have helped.

Then inventory the agents you already have. Check the CRM, the email platform, the dialler and the meeting scheduler for agent features that are switched on. Many teams will find they are further along the forecast than they thought, with nothing logged.

Finally, pick one queue and one owner. Our view: a sales agent with a named owner, a narrow permission set and the five log fields will tell you more in ninety days than ten agents switched on by default. When the owner can show a seller a faster follow-up time and a cost per meeting, the seller's answer to Gartner's question changes. Our AI automation page describes how we scope that first workflow on the CRM you already run.