AI-agent adoption is advancing faster than campaign measurement

A customer-facing AI agent can answer a prospect’s questions, qualify their interest, and move a sales conversation forward. But marketing can connect that work to advertising spending only if it can link the conversation to both the originating campaign and the eventual purchase. When the purchase stays in a separate system, marketers lose information they need to judge the campaign and decide where to spend next.

Invoca’s 2026 Broken Funnel Report shows how widely adoption has spread: 44% of business-to-consumer (B2C) marketers have conversational AI agents operating in at least one channel, and another 27% are testing them. Invoca benefits commercially when businesses invest in conversation measurement. Measurement includes returning conversion feedback, information about completed sales or booked outcomes, to advertising systems so they can use those results to improve campaigns.

To examine adoption and measurement, Invoca surveyed 451 U.S. marketers in September and October 2026 at companies with at least 100 employees. The sample covered home services, healthcare and dental, financial services and insurance, automotive, and telecom. The adoption figures concern B2C marketers; the findings below distinguish marketers using or testing agents from the wider survey population where applicable.

Campaign measurement requires three connected steps

The first measurement step is connecting an agent conversation to the advertising activity that generated it. Only 26% of marketers in Invoca’s findings could connect those conversations to an originating campaign, keyword, or ad. Without that link, even a productive exchange tells marketing little about which acquisition decision brought the customer in.

Following the customer from that exchange to a completed outcome is the second step. Nearly three-quarters of marketers using or testing agents could not track the associated leads through conversion, according to Invoca. Knowing which campaign generated a conversation is a start, but the team still needs to learn whether it led to a sale or booking.

The third step is returning the outcome to the systems managing advertising, and Invoca found that only 19% of marketers could reliably send agent conversion data back to those platforms. Advertising systems need those results to assess performance and guide future spending. Without them, media-management tools, which operate campaigns, and attribution tools, which connect results to the activity that helped generate them, work with an incomplete account.

A local-branch sale shows how these three connections can break within one customer journey. A campaign generates a lead, an AI agent handles the conversation, and a salesperson at a local branch completes the sale. The campaign team records the lead, the agent records a productive exchange, and the branch records revenue. Unless those records stay connected, each stage can appear successful while marketing struggles to determine how the earlier interactions contributed to revenue.

Because the outcome can sit beyond the agent’s immediate view, evaluating the agent requires separate questions about deployment, service, and sales. Deployment establishes that customers can use the agent; service measures show how well it handles the work; revenue and booked outcomes show the commercial result. Invoca found that only 27% primarily assessed AI performance through revenue or booked outcomes. The remaining respondents used measures such as customer satisfaction, qualified leads, conversation volume, the proportion of conversations completed without human involvement, or cost to serve.

These service measures answer different operating questions. Customer satisfaction indicates whether the interaction served the customer well, while conversation volume shows how much work the agent handles. The share completed without human involvement and cost to serve help assess workload and efficiency; qualified leads show progress toward a sale. So an agent can handle questions efficiently and provide useful service even while its contribution to completed sales remains difficult to measure.

For campaign decisions, the practical question is where the final outcome becomes visible and whether it can return to advertising systems. Answering it requires connecting the agent’s records with later customer behavior and the sale or booking. Marketing then has evidence for its next spending decision as well as a fuller account of the previous one.

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.

Agents add a handoff to an existing attribution problem

Connecting those records becomes harder when sales already cross several systems and organizations. Across Invoca’s wider survey population, 74% could not fully follow leads from their initial digital interaction through conversion. This finding concerns the broader customer journey, whereas the tracking figure above concerns marketers using or testing agents. An agent enters a process in which those connections may already be weak.

The largest reported break in that broader journey occurred when a lead reached a dealer, franchise, or local location, where Invoca found that 56% of marketers lost attribution. As in the local-branch example, this is both a sales step and an organizational handoff: marketing may hold the acquisition record while the location holds the transaction. To measure the journey, those parts of the business must keep their records connected.

The connection can also break as customers change communication channels. Invoca found that 46% of marketers lost attribution in text or chat, and 40% lost it on phone calls. Customers can stay engaged with the same business throughout these interactions while marketing loses track of the sequence. These percentages describe separate points of failure and should be considered individually.

Even when marketers can follow a journey, they may have few opportunities to influence it. Only 21% of respondents in Invoca’s survey said they could both observe and influence leads all the way to conversion. A functioning agent makes one interaction available to customers, but measuring its contribution requires information to stay connected across the surrounding calls, messages, systems, and sales locations.

AI-origin lead quality measures an earlier stage

The distinction between stages also matters when evaluating leads that begin with AI-assisted research. In Invoca’s survey, 44% of marketers identified AI answer engines, including ChatGPT, Claude, and Gemini, as supplying their most qualified leads. These answer engines help customers research a purchase before they reach a business. Company-deployed agents serve a separate role by answering questions or supporting the purchase within the business’s customer journey.

Marketers’ assessments in Invoca’s survey placed AI answer engines ahead of several established acquisition channels. Google Local Services Ads was identified by 40% as supplying their most qualified leads, followed by organic search at 39%, paid social at 37%, and paid search at 36%. These figures reflect respondents’ judgments of lead quality across the categories. The AI-answer-engine figure groups ChatGPT, Claude, and Gemini together.

Alongside those quality assessments, Invoca found that 86% of marketers said they could identify an AI-answer-engine research origin for most or some leads. Recognizing that origin gives marketing a useful starting point for understanding acquisition. Customers can then move through calls, chats, company agents, local locations, and sales conversations, and each step needs to preserve the connection to their origin.

Because that later journey still determines the commercial outcome, promising AI-origin leads and incomplete conversion measurement can coexist. Lead quality tells marketers about the prospects a channel supplies; following those prospects through a sale shows whether they become completed business. Marketing can value the first result while building the connections needed to measure the second.

Customer context and revenue attribution require different connections

Following a customer’s research also matters inside the agent conversation, where earlier activity can help the agent respond appropriately. Only 27% of marketers in Invoca’s findings said their agents could access and use a customer’s earlier digital journey. A prospect may already have researched a product or interacted with the company elsewhere. If the agent cannot access that history, it may treat a returning prospect as a new contact.

The improvements marketers requested in Invoca’s survey show what they want agents to do with that history. The most requested improvement, cited by 43%, was adapting conversations using the customer’s digital journey. Another 41% wanted agents to recognize returning customers and use their history. Both capabilities require earlier activity to be available during the interaction so the agent can respond in a way that reflects what the customer has already done.

Making earlier activity available to the agent improves its understanding of the customer, while revenue attribution requires the eventual outcome to reach marketing. An agent could recognize a returning prospect perfectly even though the final sale remains disconnected from the originating campaign. Conversely, a company could track the completed sale while its agent still lacks the history needed to serve that customer well. Teams evaluating either connection therefore need to test what information reaches its intended user: customer history for the agent’s next response, and conversion outcomes for marketing’s next spending decision.

Key highlights

  • Tie adoption to measurement readiness: Invoca found that 44% of B2C marketers have conversational AI agents operating, with another 27% testing them. Marketing and technology owners need to assess whether campaign, conversation, and sales records stay connected as deployment expands.
  • Close the conversion feedback loop: Only 19% of marketers could reliably return agent conversion data to advertising platforms. Marketing operations needs to connect originating campaigns, agent conversations, and completed outcomes so advertising systems can use sales results to guide spending.
  • Assign ownership across sales handoffs: Across the wider survey, 56% of marketers lost attribution when leads reached dealers, franchises, or local locations. Marketing and local sales owners need shared responsibility for connecting acquisition records to completed transactions.
  • Validate AI-origin lead quality through sales: Although 44% of marketers identified AI answer engines as supplying their most qualified leads, that assessment captures an earlier stage of the journey. Acquisition teams need to follow those leads through conversion to establish their commercial value.
  • Test customer context and attribution separately: Only 27% said their agents could access and use a customer’s earlier digital journey. Agent owners need to test whether customer history informs responses, while marketing operations verifies that completed sales feed campaign measurement.

Alexander Procter

November 18, 2026

8 Min

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.