What agentic AI actually means for a field marketing team

Gartner predicts that by 2028, AI agents will outnumber sellers ten to one. In the same forecast, it predicts that fewer than 40% of sellers will say agents improved their productivity (Gartner, 2025).

Both halves are predictions, not measurements. Read together, they describe an enormous amount of software doing an enormous number of things, while most of the people working next to it do not find it especially useful. That is a reasonable starting point for any field marketer trying to work out what the word "agentic" should mean for their team this year.

What vendors mean by an agent

In practice, the difference between an AI assistant and an AI agent comes down to permission. An assistant drafts the email and waits. An agent is given a goal and allowed to take multi-step actions across other systems: send the email, update the record, book the meeting, move on to the next contact.

For a field team, the question that matters is which systems the agent can write to. An agent that drafts is a writing tool. An agent that writes to your CRM, sends from a rep's inbox or messages a guest the night before a dinner is making decisions on your behalf, and each of those decisions is visible to a customer or to leadership.

Where adoption actually stands

The piloting is widespread. Eighty-one percent of martech leaders are piloting or implementing AI agents, and only 1% have no plans (Gartner, 2025). The top pilot uses are content production (52%), content enrichment (49%) and campaign management (43%). On the sales side, 54% of sellers say they have used AI agents, according to Salesforce's vendor research (Salesforce, 2026).

Commitment is much thinner. Only 3% of B2B marketers consider agents core to their strategy, even though 28% are experimenting with them (Content Marketing Institute, 2025). Scale depends heavily on size: 40% of large enterprises with more than $1 billion in revenue are scaling AI agents, against 22% of smaller organizations (McKinsey, 2026).

The platform vendors report fast uptake of their own products. HubSpot says its Breeze Prospecting Agent has nearly 17,000 customers (HubSpot, 2026). Salesforce reports $1.2 billion in Agentforce annual recurring revenue (Salesforce, 2026). Salesforce also says its own internal SDR agent contacted 130,000 previously untouched leads in four months and generated 3,200 opportunities (Salesforce, 2026). That is company-reported, and the word "untouched" is doing a lot of work: those were leads no human had followed up, which says as much about the backlog as about the agent.

The skeptic's ledger

The critical numbers come mostly from the same analysts. Forty-five percent of martech leaders say vendor-offered AI agents fail to meet expectations for promised business performance. Half say their technical or data stack isn't ready for agents, and half lack the technical talent (Gartner, 2025).

Gartner also predicts that more than 40% of agentic AI projects will be canceled by the end of 2027. It counts only about 130 of the thousands of vendors selling "agentic" products as the real thing, and has a name for the rest of the category's marketing: "agent washing" (Gartner, 2025).

The most useful Gartner line for a field marketer is a forecast about sales leaders: those who overhaul data, automation and user experience will be five times more likely to get ROI from AI than those who choose quick fixes (Gartner, 2026). An agent acting on your attendee and CRM data is only as good as that data. For event teams, whose attendance history often lives outside the CRM entirely, that is the constraint that decides everything else.

The agents on the other side of the table

Field marketers should also pay attention to agents they will never buy. Sixty-one percent of purchase influencers say their organization has, or will have, a private generative AI engine to support purchasing (Forrester, 2025). Forrester predicts 20% of B2B sellers will be pushed into agent-led quote negotiations in 2026. Gartner predicts AI agents will intermediate more than $15 trillion of B2B spending by 2028 (Gartner via Digital Commerce 360, 2025).

Those are forecasts, and large ones. The direction still matters for events. If more of the comparison and shortlisting work moves to software on the buyer's side, the moments that software cannot handle become more valuable. An agent can compare your spec sheet with a competitor's. It cannot sit at a dinner and ask your customer's head of operations what went wrong in year two.

Realistic near-term jobs for an events agent

Given all of that, the sensible near-term agent jobs for a field team are narrow, supervised and reversible.

Logistics messaging is the clearest candidate. Confirmations, reminders, calendar updates, directions and answers to routine questions such as parking and dress code are repetitive, low-stakes and easy to review. A mistake here is embarrassing but fixable.

List reconciliation is a good job for an agent that proposes and a human who approves. The agent matches registrants to CRM contacts and accounts, flags likely duplicates and suggests which new names need records. A person approves each change before anything is written. The proposing is the tedious part, and the approving is where the risk lives.

Pre-event brief assembly works the same way. An agent can pull CRM history and public information into a one-page brief per guest. A human verifies job titles and anything recent before the host reads it.

Follow-up queuing is the fourth. After the event, an agent can draft follow-ups by attendee segment and route each one to the owning rep for approval, then log the activity once it is sent.

Some jobs should stay with people for now. Deciding who gets one of 14 seats. Writing to the CRM without review. Anything that speaks in the host's voice to a senior guest. A useful rule of thumb is that autonomy should track reversibility. A reminder email can be corrected. An overwritten account record often goes unnoticed for a quarter, and a careless note to a senior executive cannot be unsent.

The modest version

For most field teams in 2026, the agent worth having does the tedious part of a job the team already does, asks before it writes anything that matters, and keeps a log of what it did. That is a modest ambition next to the keynote demos. Given Gartner's cancellation forecast, the modest version is the one most likely to still be running in 2028.