1. Prospect
A prospect fills out the form
The submission timestamp starts the clock.
2026 SPEED TO LEAD REPORT
Speed to lead
6,346
Forms tested with agents
The first GTM Engineering Benchmark from Clay. We sent AI agents to test what happens after a buyer fills out a B2B form—and how every company stacks up.
Enter your company email to see how you compare to your peers.
Create custom research agents by describing your task, connecting tools, and deploying to any table.
1. Prospect
The submission timestamp starts the clock.
2. Company
The first observed response stops the clock.
3. Measurement
Every response can now be compared with the benchmark.
AI has changed what's possible in GTM. The introduction of AI SDRs, enrichment waterfalls, and autonomous research agents has moved the goalposts for everyone at the same time. We're all learning in real-time how to fold automation into the way we sell.
This raises a question: how do you know if you're actually good at it?
That's what our GTM Engineering Benchmark series is for. Each report picks one metric that matters, measures it across the real internet, and shows you how to benchmark your level of execution against the data—and then how to move the number. We'll explain what the metric is, why it matters, what the data shows, how to benchmark yourself, how to improve, and the exact build behind it.
We're starting with speed to lead: how fast you respond when a prospect fills out a form. Few metrics tie more directly to pipeline, but our results show most companies miss the mark.
Filling out thousands of forms by hand would be time-consuming and expensive for a study. Instead, we built AI agents—orchestrated by Clay—to do the work.
Each agent went to a company's website, found the demo or contact form, and submitted it. When a company replied by email or called, the response was timestamped automatically. In total, the agents filled out 6,346 real demo and contact forms.
To do this, an agent needs the same primitives a human uses to interact with the world: a way to reason and navigate a form, an email address, and a phone number. We partnered with three startups that are providing those primitives today:
One Clay table orchestrated the entire run. The agents pulled their assignments from it, the inboxes and phone numbers reported back to it, and the analysis ran inside it. The same sequence repeated 6,346 times, and no one on our team touched a single form. Here's how:
The study starts with Clay's Tier 1 and Tier 2 accounts. Each row gives the agent a company and domain to investigate, while Clay keeps the full test set structured and auditable.
Claygent visits each domain and locates the exact contact or demo-request URL. This avoids testing generic homepages and gives the browser agent a precise destination.
Request a demo
Tell us about your team
Correct B2B form confirmed
The browser-control agent submits one consistent synthetic identity. AgentMail and Saperly provide the monitored email address and phone number used to observe every response.
Request a demo
company.com/request-demo
Name
Jack O'Brien
Title
Founder & CEO
Company
Subconscious
Work email
AgentMail email address
Mobile
Saperly mobile number
The system pauses after submission. No follow-up is sent. The clock keeps running until an email or phone response arrives—or the observation window closes with no reply.
00:37:00
Form submitted
00:00:00
First response
00:37:00
AgentMail catches email replies with their timestamps. Saperly listens for calls to the submitted mobile number. Together they cover the two response channels in the benchmark.
inboxRe: Your demo request
Incoming call
Each provider posts its response payload into a listening Clay table. Formula columns compare the original form-fill time with the observed response time to produce speed to lead.
Claygent compares timestamps, identifies human versus automated senders, and scores the level of personalization. Those structured outputs become the benchmark distributions in this report.
Speed
3h 42m
Sender
Human
Personalization
Full
If you're building agents for GTM, the ceiling on what they can do is set by which primitives they can reach. Give an agent an inbox and a phone number, and it stops being a chatbot and starts being something that can act in the real-world. GTM agents will need these primitives as the future of GTM goes from agents to orchestrated agents to self-optimizing systems from learning loops.
Structured benchmark data, ready to analyze
Like any study of this size, ours has limits. These are three worth knowing before we dive into the data:
Synthetic buyer identity
Same identity on every form
| Metric | Result | Methodology |
|---|---|---|
| Form complexity | Number of fields | Claygent analysis |
| Response | Yes or no | Whether or not we received an email or call |
| Speed of email follow up | Time in minutes, hours, or days | Compared the submission timestamp with when a follow-up email reached our AgentMail inbox |
| Speed of phone call follow up | Time in minutes, hours, or days | Compared the submission timestamp with when our Saperly phone number was called |
| Personalization level | Template, light, or full | Claygent categorization |
| Who sent it | Human vs. automation | Claygent AI estimate based on timing and personalization |
| Correct qualification | Yes, no, or unsure | Claygent analysis of the website ICP compared with the identity we submitted |
The data set
Get the benchmark and compare your speed
Open your personalized benchmark and see where your company stands.
2026 Speed to Lead Benchmark — one metric that matters, measured across the real internet.
© 2026 Clay Labs, Inc.
Terms · Privacy · Security