What is Convin & Rule Engine?
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
In late 2023, Convin was preparing to onboard a major enterprise client but, we hit a wall.
Their QA workflows weren’t just complex; they were deeply conditional, multi-layered, and full of nested exceptions:
Delivery Logic: Redelivery after 3 failed attempts, real-time dispatch validation, stock checks
Incident Logic: Split flows for lost vs. damaged parcels with unique refund/escalation conditions
Our existing Python-based auditing system couldn’t scale. Custom logic for every template took 7–14 days, required engineers, and wasn’t reusable.
That’s when I proposed building something bigger:
A visual, self-serve rule builder for internal and external users.
A system that could eventually power not just audit logic, but AI agent workflows (e.g., voice agents, chat automations).
A way to reduce internal ops headcount and boost product adaptability across verticals.
This wasn’t just about landing a client. It was about turning a service into a product.
Research & Discovery
Since we were short on time, I did a deep dive into our internal process:
Met with our Product Delivery Team to understand why each template took days.
Shadowed engineers building Python rules line-by-line for 3–8 templates, each with 10+ parameters.
Reviewed past delivery tickets to spot common bottlenecks.
Insight: Most complexity stemmed from lack of modularity and low traceability in the system
On the market side:
I researched competitors — but none offered advanced logic beyond basic AND/OR flows.
So I zoomed out and looked at Salesforce, Zapier, Webflow, and internal tooling from other industries.
That’s when I remembered my days at IBM, seeing Salesforce’s workflow engine: powerful, but painful to use.
This gave me the conviction: Convin needed a real rule engine, not a logic toy. It needed:
Reusability
Nested logic support
Testability
Transparency for QA leads and non-technical users
Design Exploration & Iteration
The initial design was drag-and-drop — a classic visual builder. But during internal testing, we hit real usability friction:
Most users on Windows ran browsers at 125–150% zoom → canvas felt too tight → drag actions became unreliable
So we pivoted:
Removed side-panel drag/drop.
Replaced with modal-based “choose block” flow.
Allowed for single block or group creation to keep things modular and visible.
We explored:
Inline vs. tray-based block previews.
Collapsible rule cards for long flows.
Hierarchical nesting vs. same-level groups.
The result: a cleaner interface that adapted to most screen sizes, even if it meant sacrificing some visual flair.
What Didn’t Work (Failures & Tradeoffs)
I initially assumed this would be quick to build — but that was naïve.
It took us over 120 days to build and stabilize what became a full-fledged product — not just an MVP.
Our backend team needed time to architect a system that could parse, store, and execute logic blocks from a UI layer.
At the time, I didn’t fully understand the technical term "rule parsing" — but I learned on the go. I realized each visual block needed to map to an underlying JSON structure that our backend could process and validate.
This taught me:
How to think in terms of system-state trees and execution flows.
How to balance design abstraction with data integrity.
How logic nesting and modularity affect backend parsing and error handling.
Stakeholder conflicts:
Product wanted to launch faster with fewer features.
I had to push back: “If we ship this raw, we’ll create a new bottleneck — not a solution.”
In the end, we chose to:
Delay a few weeks to build reusable blocks and test/debug views.
Cut some complex features (currently in-progress).
Prioritize long-term scalability over short-term polish.
Governance & Scalability Considerations
We didn’t just build this for now — we’re building it for the future of Convin’s platform.
Ongoing discussions:
Rule versioning and rollback (in progress).
Audit logs per rule execution.
Templates for reusability.
Potential to power Voice Agent flows and AI Agent decision trees.
The Rule Engine has now evolved into its own mini-platform within Convin; a framework to build logic for any AI system, not just auditing.
Leadership & Influence
Even though I was the solo designer here, I acted as a Design Lead in spirit:
Mentored our visual designer to think systematically.
Guided an intern who now owns Rule Engine iterations full-time.
Conducted internal team walkthroughs for adoption.
Helped onboard clients by conducting live training calls.
Internally, I pushed back on:
Scope reduction that would hurt usability.
Dev choices that affected UX traceability.
Copy decisions that lacked clarity for QA users.
The product has seen early adoption across multiple clients, with new onboarding flows, contextual help, and self-serve tutorials underway to improve usability and scale.
Learnings
No-code = designing for invisible complexity.
Design is not just user-first, it's system-aware and constraint-conscious.
Not everything can be simplified but everything must be explainable.
Real UX decisions happen in tradeoffs, not in tools.
Building a platform? Design for the next 100 templates, not just today’s 5.
TL;DR Summary
Project: Rule Engine – No-Code Builder for Auto QA
Role: Product Designer + Strategy Partner
Skills: UX Strategy, Platform Thinking, Dev Collaboration, Team Enablement
Impact:
Reduced onboarding setup time by 50%.
Lowered engineering dependency by 70%.
Evolved from audit tool → logic platform for AI agent control.
Led cross-functional strategy from research → design → adoption.
