SYBILL CASE STUDY

How Sybill rebuilt a broken funnel into qualified pipeline machine

Sybill’s product listens to thousands of sales conversations a month. The one conversation it wasn’t reading was the one happening on its own homepage, where their chat agent was quietly turning away the enterprise buyers Sybill had grown to want. In a single quarter, Docket rebuilt the homepage into a qualification engine Sybill trusted enough to run its own inbound through.
1.8% → 18
A near-dead conversion rate rebuilt into 18 captured leads in a single quarter
Multiple/mo
Deals now sourced through Docket and attributed cleanly in HubSpot
Right Routing
Previously, strong leads were sitting unacted upon; the agent now surfaces and routes properly
The Challenge

The previous agent was perfect, for a company that no longer existed.

When Sybill first switched on its own website agent, the company was focused on founder-led sales. A tiny ICP,  and a founder’s calendar to protect. So the agent was built to be a careful gatekeeper, and for that moment it was right.
Then Sybill grew, it hired a Head of Sales and a Head of BD, standing up a real enterprise motion. But the agent kept gatekeeping for the old world, and nobody noticed how much it was costing until Docket’s CS team pulled four weeks of transcripts and showed the sales team what they were missing.
"Wait, why are we not talking to [Enterprise Company]? Give me the lead.”
-Kate Garvey, Head of Sales, Sybill
Implementation

Four cracks had opened in the funnel, and a pipeline was draining out of each one.

  1. Build the Knowledge Foundation
    The team connected the AI Marketing Agent to the main marketing website and 532 pages of technical documentation — product guides, API references, compliance documentation, and integration specs. No engineering tickets. No months-long implementation. The agent read from the same sources a solutions engineer would reach for on a customer call.
  2. Iterate in Days, Not Weeks
    As early conversations surfaced gaps, the team expanded the knowledge base with an additional 223 pages of supporting guides. The knowledge layer grew with what buyers were actually asking — iteration happened in days, not weeks.
  3. Connect Salesforce for Automatic CRM Sync
    Every conversation outcome — qualification status, intent signals, use case details, contact information — landed in the pipeline automatically. No manual entry. No blank fields on the rep's end.
  4. Enable Spanish & Portuguese from Day One
    With LATAM already showing up as a significant traffic segment, the agent was configured to engage buyers in their language before the first month of data was even in.
1
It was screening out the buyers Sybill now wanted
A live enterprise buyer mid-conversation with the agent, came back only partially qualified. The agent was still enforcing a founder-era ICP the company had long outgrown.
2
It answered everything and qualified no one
Visitors got accurate answers, then left. Built to respond and never to qualify, the agent created no intent signal, so the sales team never knew they’d been there.
3
An AI sales company was  running a help desk
With the value-led welcome gone, the agent was easy to mistake for support. The transcripts filled with “I can’t log in” and “how do I reset my password.”
4
Invisible leads, none  reached the CRM
HubSpot wasn’t fully connected, so leads were never attributed to the agent. The pipeline was leaking invisibly, and a program no one can measure is a program no one defends.
HOW DOCKET SOLVED IT

Five changes that turned a gatekeeper into a qualifier.

Sybill’s whole product is built on understanding sales conversations, so its team set a higher bar than most. Recommendations weren’t enough; changes had to be proven. So Docket ran the relaunch as a discipline that suited skeptics: change one thing, measure for four weeks, attributed every result.
“I will test and iterate anything, it can even go against my gut feeling. But test, iterate, improve, test, iterate, improve.”
-Kate Garvey, Head of Sales, Sybill
It was no longer a chatbot, it was an agent rebuilt around the company Sybill had become.
01
Re-qualified for the company Sybill had become
Criteria rebuilt around Sybill’s current team, not its founder-led past. Hard disqualifiers narrowed to genuine non-fits, and Sybill’s HubSpot target accounts wired in to fast-track named logos.
IMPACT
Enterprise-shaped buyers reach a real conversation, not a dead end. Every completed chat is an AQL.
02
Rebuilt the opening, and earned the email
A value-led opening replaced the passive welcome, stating what the agent does and asking whether the visitor is exploring or evaluating. The email-ask moved to after the agent proves its worth.
IMPACT
Intent surfaces in the first exchange, the support-desk confusion disappears.
03
Put Humans on warm leads in real time
Slack routing now pings the Head of BD the instant a warm lead is mid-conversation, so she can step in while the buyer is still on the page. Support questions were split into their own channel.
IMPACT
A high-value prospect meets a human while intent is hot, not a day later.
04
Armed it to win the comparison
Battle cards against Gong, Fathom, and Fireflies were loaded in, plus the Storylane tour, with every answer drawn only from approved material via Docket’s Sales Knowledge Lake™. No improvisation on competitive or pricing claims.
IMPACT
Competitive chats end with Sybill ahead; meeting-resistant buyers still self-select in.
05
Wired attribution to HubSpot, then ran the agent like an experiment
Every booked demo is credited to the agent in HubSpot via UTMs and first-party tracking. Nothing shipped on faith: one change at a time, four weeks of data each, every shift tied to the prompt that caused it.
IMPACT
No more “we changed some things and the pipeline moved.” Every gain is provable and repeatable.

Testimonial

The qualification changes have definitely helped. I’m seeing strong accounts come through now instead of getting filtered out for no reason. And anybody that books a demo after going through Docket, that’s all tracked.
Edie Shimel
Growth Ops Lead
THE RESULTS

A funnel that finally works for the company Sybill is today.

One outcome the team didn't fully anticipate: the agent became a direct window into what the market actually wanted to talk about. Across 532 conversations, five themes dominated — not survey data or analyst research, but real buyer intent captured in real time, in their own words.
1.8% → 18
A conversion rate that had cratered near 1.8% rebuilt into 18 captured leads in a single quarter.
4x
Conversations run 4x deeper than the pre-relaunch baseline, with buyers staying in real dialogue.
Multiple / mo
Deals now carry Docket as their booking channel in HubSpot, the first time Sybill could trace a website conversation to a sourced deal.
Right Routing
Previously, strong leads were sitting unacted upon; the agent now surfaces and routes properly
Sybill builds AI for revenue teams. They could have built this themselves, but they chose Docket instead.
They held Docket to a bar most buyers never set. Ran every change as an experiment, attributed every result. Then put their own pipeline on the agent, and their founder referred the next customer. The roadmap is still running: avatar, post-value email capture, and a voice A/B test, each shipped one experiment at a time.
SEE IT ON YOUR SITE

SEE IT ON YOUR SITE

If the hardest audience to convince is already convinced, the question isn’t whether the agent works. It’s how much pipeline you’re leaving on the floor until it does.
Book a demo
By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts.
To learn more, check out our Cookie Policy here.
PreferencesDenyAccept