Role: Sr. Manager, UX
Defining a CX modelling system for complex customer conversations
A system that enabled teams to design and test conversational CX workflows, simulating real customer interactions before implementation.
FOCUS: Product Strategy | CX Architecture | Workflow Architecture | Cross-Functional Alignment
Product context
Platform: Cyara Velocity
Domain: CX Testing & Conversation Design
Contact centres were increasingly automating customer requests across voice and chat channels.
Designing these experiences required navigating fragmented configuration tools, scripts and system interfaces. Conversation designers, product managers and engineers often worked across different representations of the same customer journey.
As conversational systems grew more complex, teams struggled to maintain a clear model of conversational paths, making it difficult to simulate and test CX behaviour reliably.
The opportunity was to create a visual CX modelling system that structured these conversations and aligned with the underlying platform architecture.
Key insight
Conversational experiences are not simply flows. They are decision systems that determine how customer requests are interpreted, routed and resolved.
Traditional flows often hid this complexity because the logic driving the experience lived across configuration layers.
The design challenge became:
How might we create a visual CX modelling system that represents conversational logic clearly enough for teams to simulate and test customer interactions?
Early exploration mapped how conversational systems combine intent recognition, decision logic and modular CX paths. Click the image for full view.
Product decisions
1. Visual CX Model as the source of truth
We designed a visual CX modelling system where conversational experiences were defined as structured models rather than disconnected flows.
​
As part of my role in Product Management, I led all aspects of the visual and interaction designs for the creation and enhancements of CX Models.
The CX Model evolved in two stages, from traditional IVR flow design to conversational AI orchestration.​
​
CX Model 1.0 - Foundational IVR orchestration
-
Dynamic and static prompts
-
Speech and DTMF menus
-
Telephony events (transfer, hang up, queue)
-
Multi-channel linkages
CX Model 2.0 - Conversational AI orchestration
-
Intent recognition through NLU
-
AI-driven conversational routing
-
Dataset-driven node configuration
Click the image for full view.
Why it matters:
A shared CX Model engaged product, design, and engineering teams in a collaborative process, aligning stakeholders on the precise behaviour of customer journeys before implementation and testing.
2. Decision nodes for deterministic routing
Decision nodes represented structured logic conditions that determined how a conversation should proceed.
These nodes allowed teams to route interactions based on:
• User input
• Customer attributes
• Contextual conditions
Click the image for full view.
Why it matters:
Made routing logic visible and easier for teams to simulate and validate CX paths.
3. Dataset-driven validation
Some conversational decisions were driven by large datasets such as:
-
Customer profiles
-
Product catalogues
-
Eligibility rules
​
Instead of embedding complex logic directly in nodes, we introduced dataset-driven validation.
Datasets could be connected to decision nodes to dynamically route conversations based on structured data inputs.
Click the image for full view.
Why it matters:
This limited complexity in the visual model while expressing powerful business and routing logic.
4. NLU nodes for intent recognition
Natural Language Understanding (NLU) nodes represented points where the system interpreted user input through trained intent models.
These nodes allowed designers to define:
• Intent recognition boundaries
• Fallback behaviour
• Confidence thresholds
Click the image for full view.
Why it matters:
Separated intent interpretation from routing logic, making conversational paths robust, easier to model and test.
5. Modular CX modelling for collaboration
Large conversational systems often involve many contributors and required teams to break CX models into manageable modules.
To support collaboration, the modelling system introduced modular CX structures that allowed teams to compartmentalise different sections of the experience.
​
Modules allowed teams to:
-
Isolate specific conversation segments
-
Reuse common interaction patterns
-
Maintain clarity in large CX models
Click the image for full view.
Why it matters:
This enabled cross-functional teams to collaborate on a single CX model without overwhelming complexity.
Business Impact
Expanded the audience for CX design
The CX Model made conversational logic visible and understandable beyond engineering teams. Business leaders, CX specialists and VUI designers could now understand and participate in shaping CX flows rather than relying solely on engineers to interpret IVR logic.
Enabled self-service CX authoring
By representing conversational logic through a visual model, teams could design and iterate on CX flows directly. This reduced dependency on engineering teams and allowed CX specialists to independently refine customer journeys.
Strengthened product positioning
The CX Model provided a clear way to demonstrate the platform’s capabilities during sales and customer engagements. Complex conversational systems could now be explained visually, helping prospects quickly grasp the value of the Velocity platform.
Reframed the product’s value
The modelling system repositioned Velocity beyond a testing platform into a strategic CX design tool, helping organisations see conversational experiences as systems that could be designed, governed and optimised across teams.
