What did Invoca's Broken Funnel Report find?
Invoca surveyed 451 US consumer-brand marketing leaders and found that 74% of those running AI agents can't track a lead the agent handled from the first ad click to a sale. Only 27% measure their agents on revenue, bookings or conversion rate. The report came out on November 10, 2026.
The agents in question are the chat, text and voice bots that now answer inbound leads. According to the press release, seven in ten of the brands surveyed have agents live or in pilot. That's 323 of the 451 respondents, and the measurement numbers apply to that group.
Invoca's CEO Gregg Johnson put it this way: "Too many consumer brands measure their AI agents like a help desk and expect them to perform like a sales team."
How are companies measuring their AI agents instead?
Mostly on activity. Customer satisfaction is the top primary measure at 20%, followed by qualified leads and conversation volume at 16% each. Only 7% of agent-running marketers measure their agents primarily on revenue itself, according to PPC Land's breakdown of the full report.
| Primary measure for AI agents | Share of marketers running agents |
|---|---|
| Customer satisfaction | 20% |
| Qualified leads | 16% |
| Conversation volume | 16% |
| Conversion rate on agent conversations | 11% |
| Containment (resolved with no human) | 10% |
| Cost per contact | 10% |
| Appointments booked | 9% |
| Revenue from agent conversations | 7% |
The tracking gaps run deeper than the scorecard. Only 26% can tie an agent conversation to the campaign or ad that produced it, and only 19% can reliably send the outcome back to their ad platforms. Once a lead is handed to a dealer, franchise or local branch, 56% lose attribution.
Why does it matter that only 19% send results back to the ad platforms?
Google and Meta bid for you based on the conversions you report. If your AI agent books a job but the ad platform never hears about it, the algorithm keeps optimizing for whatever signal it does get, like a form fill or a chat started, and it buys you more of those.
PPC Land spells out the risk. A brand's agent can convert more leads while the bidding system learns nothing from those sales. Invoca's own release warns that AI improvements in sales effectiveness "may inadvertently hamper marketing efficiency." Your ads end up chasing people who chat and never buy.
The home services numbers are worse. In Invoca's home services edition (137 respondents), 81% can't track an agent-handled lead to a conversion and only 14% send outcomes back to their ad platforms.
Is this survey reliable?
It's useful, with limits. It's self-reported, US only, and covers companies with 100 or more employees in five industries. Invoca also sells call tracking and AI agents, so the problem it found is one it sells a fix for. Treat the numbers as a rough direction.
The full report gives a margin of error of plus or minus 4.6 points across all 451 respondents, so the band is wider on the 323 running agents. The survey also doesn't compare how agent-handled leads convert against leads handled by people.
What should a business owner who runs ads and uses AI agents do?
Score your AI agent on booked appointments and closed revenue, and make sure every sale it starts is traced back to the ad that paid for it. If you can't connect a closed deal to a campaign, you can't tell which ads are making you money.
- Pick one number for the agent. Booked appointments that showed is a good one. Conversation count is a bad one.
- Tag every lead with its source the moment it arrives, before the agent says a word, so the campaign stays attached through the call, the booking and the sale.
- Mark deals won or lost in one place and send those results back to your ad platform. Meta and Google both accept offline conversion data.
- Once a month, put your revenue from closed deals and your real ad spend into our free ROAS calculator. If the result doesn't match what your ads dashboard says, your tracking is broken somewhere.
I've been running ads since 2013, and the accounts that scale are the ones where every lead has a dollar figure next to it at the end. In KenjiAI, a Voice AI agent or the chat agent answers the lead, and the conversation, the booking and the payment all land on the same contact in the CRM pipeline. You can see which ad brought in which sale without exporting three spreadsheets.
This week, pull ten leads your AI agent handled last month and try to follow each one to a sale or a dead end. Count how many you lose track of. That's your own version of Invoca's 74%.