Invoca报告：营销人员部署AI智能体速度远超其衡量归因能力

Marketers are putting AI agents in front of customers faster than they can connect those conversations to revenue.

Forty-four percent of B2C marketers already have conversational AI agents live in at least one channel, and another 27% are testing them. Yet only 19% can reliably send agent conversion data back to advertising platforms, according to Invoca’s “2026 Broken Funnel Report.”

That leaves marketers with a familiar measurement problem in a new place. The agent can answer questions, qualify a prospect, or move someone toward a sale, while the systems that manage media and attribution may have little idea of what happened.

Nearly three-quarters of marketers using or testing agents can’t track those leads from click-through conversion. Only 26% can tie agent conversations back to the campaign, keyword, or ad that generated them.

The disconnect helps explain how companies are judging AI performance. Only 27% use revenue or booked outcomes as their primary measure. The rest rely on metrics such as customer satisfaction, qualified leads, conversation volume, the share of conversations handled without a person, or cost to serve.

Those measures can show whether an agent is busy or efficient. They say considerably less about whether it produced business that would justify the investment.

The customer journey disappears before the sale

AI agents sit inside a broader attribution problem.

Seventy-four percent of marketers surveyed can’t fully track leads from the first digital interaction through conversion. The biggest break comes when a lead reaches a dealer, franchise, or local location, at which point 56% lose attribution. Forty-six percent lose it in text or chat, and 40% lose it on phone calls.

Only 21% say they can both see and influence leads all the way through conversion.

AI adds another handoff to a journey that was already difficult to connect. A campaign can generate the lead, an AI agent can handle the conversation, and a local branch or salesperson can close the sale. If those systems don’t share data, each part of the journey can look successful, without marketing knowing which interactions contributed to revenue.

The gap reaches customer experience as well. Only 27% say their AI agents can see and use a customer’s previous digital journey. An agent without that context may treat an existing prospect as new, even if the person has already researched the product or interacted with the company elsewhere.

Marketers appear to recognize the limitation. The most requested improvement for AI agents was the ability to tailor conversations using the customer’s digital journey, cited by 43%. Another 41% wanted agents to recognize returning customers and use their history.

AI search is creating better leads 

The shift starts before customers reach the agent.

Forty-four percent said AI answer engines such as ChatGPT, Claude, and Gemini produce their most qualified leads, ahead of Google Local Services Ads at 40%, organic search at 39%, paid social at 37%, and paid search at 36%.

Most marketers can identify at least some leads that began their research in an AI answer engine. Eighty-six percent said they could do so for most or some leads. The visibility fades as those customers move through calls, chats, agents, local locations, and sales conversations.

That creates an awkward setup for media optimization. AI can influence the research, and an AI agent can handle the next conversation, and the eventual sale can happen somewhere else. If the conversion never makes it back to the systems buying media, campaigns are still being optimized on a partial view of what worked.

The AI agent race has moved past deployment. The harder work is connecting what the agent knows, what the customer does next, and what ultimately produces revenue.

Invoca’s findings come from a September and October 2026 survey of 451 U.S. marketers at companies with at least 100 employees across home services, healthcare and dental, financial services and insurance, automotive, and telecom. The full report can be found here. (No registration required)

The post Marketers deploying AI agents faster than they can measure them appeared first on MarTech.