Build a Unified, Trusted AI Assistant
Evolve Sybill beyond a simple chat box to integrate intelligence seamlessly into sales workflows
Overview
Sybill had made its name as an AI meeting notetaker. Ask Sybill was its next bet: an assistant reps could just ask, in plain language, to pull insight from all their sales data and act on it. The hard part was making a net-new AI surface people would discover, trust, and return to, then keeping it ahead as models and expectations moved fast.

I owned the end-to-end design of Ask Sybill across its life, from the 0-to-1 build to a major upgrade a year later, growing it from a single-question box into an assistant reps rely on daily and moving Sybill from notetaker toward a true sales assistant.
My Role
Sole designer leading end-to-end design
Time
2025
Team
Product Manager
Frong End Engineer
Back End Engineer
AI Engineer
Duration
Beta & GA - 2 months
V2 upgrade - 1.5 months
my contribution
🔮  Defined a net-new AI assistant from zero: presence, entry points, and the full chat experience.
🧢  Led a major upgrade end to end, stepping into the PM role to scope and sequence with engineering.
🎨  Set foundational AI chat patterns that Sybill's later AI features built on.
🚀  Turned research on trust and adoption into design bets that deepened daily usage.
impact
🎉
3.8K users, 40K+ queries
Ask Sybill reached broad adoption across Sybill's customer base as a net-new AI surface.
🎉
64% saved 30+ minutes a week
Reps cut real time out of call prep, deal research, and synthesis across meetings.
🎉
Over 80% would recommend it
Strong advocacy from AEs, with 56% giving a perfect 10.
🎉
Moved Sybill from notetaker to daily assistant
The feature extended Sybill from summarizing meetings into the rep's everyday selling motion, anchoring its shift toward a true sales assistant.
*Metrics reflect the beta release, prior to the V2 upgrade.
CONTEXT
Turning scattered sales data into answers reps can act on
A sales team's calls, emails, and tools hold real insight, but it's scattered across the CRM, Slack, and other tools, hard for reps and managers to reach. This raises a question:
❓
Could we create a tool that lets people access all this data and derive insights through a simple solution?
With ChatGPT making conversational AI familiar, a chat experience was the natural fit: easy to adopt and quick to get value from. So we built Ask Sybill, an AI assistant you pull insight from just by asking, the way you'd ask a colleague.
To validate the need, we started small. We added AI chat entry points on the single meeting and single deal pages, so a rep could ask questions about that one entity and get answers back.
When demoing this concept to customers, we've observed fascinating use cases — customers have shown interest in exploring virtually any topic through this interface. So we decided to fully invest in opening it all the way up, which meant building the whole experience from zero.
🎯
Ask Sybill project goals
  • Expand the single Q&A experience to a multi-turn AI chat experience
  • Expand the chat capability from single meeting/deal to asking across multiple meetings and deals at once
  • Kick off sales collateral like a follow-up or mutual action plan with AI artifacts initiated in the chat.
With Ask Sybill, that scattered data finally becomes reachable. Instead of rewatching calls or digging through summaries and CRM fields, reps just ask, and turn the answers into sharper insight and faster action on their deals, which is what closing faster comes down to.
PHASE 1
Building Ask Sybill from 0 to 1
Most customers still saw Sybill as an AI notetaker for sales meetings. Phase 1's goal was to introduce and integrate the Ask Sybill experience into the app.
Define the AI chat presence
Discoverability and activation are the first step to adoption, which made the AI's presence and entry points critical to get right. I ran extensive market research and explored a wide range of options by making some early designs.
After reviewing the explorations with the team, we started with 2 entry points:
1. A new tab in the Nav bar
2. A chat bar sits at the top of each Sybill page (Meetings, Deals, single meeting, single deal
Both placements followed the same principles:
Define the chat experience foundation
An entry point is only as good as what happens after someone asks. So the larger half of this phase was the layer underneath: how a conversation and its canvas behave, how a user follows the AI's reasoning, and how they trust the result. I benchmarked AI tools on the market for inspiration, then designed the pieces that make up a full chat experience:
Based on feedback from user testing, I also added functionalities that make asking AI questions easier:
PHASE 2 • V2
Making Ask Sybill more delightful and powerful to use
8 months in, two things made it time for a major upgrade.
Three frictions were capping adoption
Since the team committed to a major upgrade of Ask Sybill, I ran evaluative research on the live experience, watching how reps used it and gathered feedback. 3 major frictions stood out:
🎉
Answers lived away from the work
Chat opened on a separate page, pulling reps out of the meeting or deal they were in, so the assistant felt like a detour rather than part of the flow.
🎉
Lacked capabilities reps now expected
As AI tools matured, reps came expecting to upload a file or image, pull in live web results, add context without spelling it out, and pick a faster or deeper thinking model. Ask Sybill couldn't do those things, so it felt a step behind others.
🎉
Reps couldn't tell what Sybill saw
It wasn't clear what Sybill knew, what context it drew on, or how it reached an answer. With no citations or view into its reasoning, that uncertainty ate at trust, and reps who didn't trust an answer stopped coming back.
My challenge: Leading the upgrade end to end with PM’s work
At the time, our PM's bandwidth was committed elsewhere, so after bringing him the research and aligning on direction, I stepped in to run the project myself.
The scope was big and may spread across multiple releases, so I built a prioritization spreadsheet to sequence features with the Head of Engineering, weighing user impact against effort, and used it to track progress at every checkpoint so nothing slipped between releases. In practice I took on the PM's role for this work, owning scope, priorities, and timeline alongside the design.
SOLUTION
Rebuilt around how reps actually work — A side panel embedded within context
The upgrade answered each of the three frictions directly. A new side panel brings the full conversation right beside the meeting or deal a rep is in, so they never lose sight of the work while they chat. A floating button lets them open, close, or reopen the chat whenever they need it.
Full clarity that enhances trust with AI
Every chat now showed the context it launched from, citations in answers, and better-exposed reasoning through a refined thinking state. Reps now have full clarity on how Sybill AI thinks and works.
A more powerful chat bar
An upgraded chat bar gave reps direct ways to shape each answer: a context selector, file and image upload, web search, and a model picker. Getting a strong answer no longer meant spelling everything out in the prompt.
outcome
From notetaker to an assistant reps rely on
The product started life as a meeting notetaker. Ask Sybill turned it into a place reps come to think through their deals and act on them, in a smoother, more capable experience, spanning everything from deal risks and competitor positioning to coaching and drafting the collateral that follows.
💖
“Just one prompt, and I get a detailed summary of all meetings from the past week along with positive and negative highlights. It's an absolute game changer. Thanks so much for making my life easier, Sybill team!”
— Leo Goldfarb @Albato
next step
Like what you see?
📩 Contact me for more project details.
Want to check out more projects?