What is Convin & Real-time Agent Assist
About Convin
Convin is an AI Agent Platform designed to optimize sales, customer service, and collection teams through three powerful agents:
AI Conversation Intelligence Agent → Analyzes customer-agent conversations across channels, audits calls, and delivers insights.
Live Agent Assist → Offers real-time prompts, objection handling, and resource support during calls.
AI Voice Agent → Handles outbound calling tasks autonomously.
The Rule Engine lives inside the Conversation Intelligence Agent, acting as the brain behind automated auditing — defining what should be checked, how, and when, all without code.
Why did we build it?
Strategic Context & Opportunity
At Convin, we were known for post-call analysis — our platform could tell managers what went wrong after the damage was already done. But what if we could help agents fix those mistakes in real-time?
That question sparked the creation of Live Agent Assist — a real-time agent support tool that activates the moment a call begins. Alongside it, we built Supervisor Assist — enabling managers to listen in, monitor KPIs, and nudge agents live, without disrupting customer experience.
The idea was initially seeded by our CEO after in-depth market research. All major competitors had something in this space — but none felt cohesive or intelligent enough.
We didn’t want to just build a checklist. We wanted to:
Empower agents to handle live calls confidently
Reduce errors, friction, and dependency on tribal knowledge
Allow supervisors to course-correct without barging in
Eventually integrate this into our broader AI Agent roadmap
What is Live Agent Assist?
Live Agent Assist is an on-call productivity layer that sits alongside agents during every call. It surfaces the right support, at the right time, without distracting the flow.
Key Features:
Call Parameter Checklist – personalized to every call type (e.g., refund flow, product demo, complaint resolution)
Real-Time Prompts – soft nudges like “Try slowing down” or “Show empathy” when needed
Battle Cards (Flashcards) – triggered based on live keywords (e.g., if the customer says “refund”, the agent sees the refund process card)
Knowledgebase Search (AI-Powered) – quick answers from uploaded docs (via our LMS system)
Closed Captions – live speech-to-text for noisy environments or difficult accents
What is Supervisor Assist?
A companion tool built for supervisors or QA managers — it lives inside the Conversation Intelligence Agent and helps:
Monitor active calls across agents
View real-time sentiment, talk ratio, adherence
Send in-call messages to agents privately (appears in the Agent Assist UI)
View live transcript and metrics without interrupting the call
Trigger Moment
Previously, our insights were always post-facto:
“You missed the refund policy here.” → Too late. Damage done.
With Live Assist, we could shift from retroactive analysis to proactive guidance. The CEO proposed it after seeing competitor movement — and we validated it further by speaking to client-side supervisors and agents. Many managers told us:
“I wish I could whisper in their ears during the call without interrupting it.”
We built exactly that.
Research & Discovery
We:
Conducted discovery calls with clients, agents, and managers
Shadowed live agent workflows across multiple verticals
Did a deep competitive and secondary research scan
Identified key friction areas: latency, distraction, layout clutter
We learned:
Agents were overwhelmed with static dashboards
Latency broke trust — if a suggestion came 5 seconds late, it was useless
Supervisor intrusion negatively impacted customer experience
UX Challenges & Design Thinking
Agents operate under pressure. In our early research, we uncovered key usability issues:
Cognitive Overload – Agents struggled when too many prompts appeared at once.
Prompt Blindness – Text-only notifications were ignored or missed mid-call.
Cluttered Layouts – Dashboard-style designs added stress rather than relief.
We realized we weren’t designing a product for browsing — we were designing for split-second, in-the-moment assistance.
Core UX Decisions & Principles:
One Nudge Per Moment: We capped real-time prompts to 1 active message at a time. If multiple triggers fired, the system queued them based on urgency.
Progressive Disclosure: Only surfaced relevant tools (e.g., KB, battle cards) based on live intent — never all at once.
Motion for Visibility: Switched from static prompts to subtle micro-animations, improving agent reaction time by ~2x (internal test).
Non-Intrusive Layout: Designed a right-side tray that stayed visible without blocking the CRM tools agents used.
Each iteration focused on reducing friction, boosting trust, and respecting the mental bandwidth of human agents.
We validated our direction through:
A/B Testing two versions with real client calls
Clickable high-fidelity prototypes tested with in-house CS teams
Supervisor feedback loops to ensure nudges were contextually accurate
Tech Constraints & Breakthroughs
We initially faced 4–8 seconds latency in closed caption and prompt triggers due to:
Call audio → Telephony provider → Convin backend → Agent UI
This broke the experience.
Breakthrough:
By introducing a recording mechanism directly into Live Assist, we bypassed some of the external latency points and reduced it to <1 second.
That change alone unlocked the entire product’s viability.
Business Strategy Shift: From Analytics Vendor → Performance Partner
Historically, Convin was seen as a “reporting tool” — we helped clients understand what happened after the call. But that limited us to QA and coaching teams.
With Live Assist, we unlocked:
Real-time influence over outcomes, not just diagnostics
New buyer personas — Ops leaders, Support VPs, RevOps, not just QA
Upsell motion from conversation intelligence → full agent productivity suite
This strategic shift expanded our TAM from post-call QA (est. $200M) to real-time enablement ($3–4B+).
It also strengthened our platform story:
Supervisor Assist tied into existing dashboards
Knowledgebase connected with our LMS product
Live nudges could be customized using our Rule Engine
Instead of a siloed tool, Live Assist became the activation layer for multiple parts of Convin’s ecosystem — and that’s why it mattered.
My Role & Leadership
Acted as the core UX owner, while mentoring a junior designer
Joined all stakeholder syncs across product, tech, and business
Partnered with PMs on roadmap tradeoffs and cut decisions
Worked closely with clients for live feedback and pilot reviews
Contributed to demo readiness and onboarding materials
