( 2. )
( 3. )
( 2. )
Chatflow Builder
Defining How Support Conversations Behave
Designed a system that defines how support conversations behave, enabling teams to scale automation and AI across channels.
Designed a system that defines how support conversations behave, enabling teams to scale automation and AI across channels.
As support teams adopted automation and AI, conversation behaviour was shaped by routing rules, agent actions, and channel-specific logic. This made interactions inconsistent and difficult to scale.
I designed Chatflow Builder, a no-code system that centralises conversation logic into a shared layer, allowing teams to define routing, escalation, and AI behaviour in one place.
This enables consistent behaviour across channels and reduces reliance on headcount as support scales.
As support teams adopted automation and AI, conversation behaviour was shaped by routing rules, agent actions, and channel-specific logic. This made interactions inconsistent and difficult to scale.
I designed Chatflow Builder, a no-code system that centralises conversation logic into a shared layer, allowing teams to define routing, escalation, and AI behaviour in one place.
This enables consistent behaviour across channels and reduces reliance on headcount as support scales.
Team
Product,
Engineering,
Leadership
Role
Product Design,
Design System
Year
2024
Problem
Support systems were built around tickets, routing rules, and agent actions, but conversation behaviour was never explicitly defined.
In practice, behaviour emerged from a combination of rules, configurations, and manual handling. This made it difficult to predict how conversations would unfold, especially as automation and AI became more involved.
Small changes could have unintended effects, and behaviour varied depending on where conversations started or who handled them. This made scaling support unreliable and hard to control.
Support systems were built around tickets, routing rules, and agent actions, but conversation behaviour was never explicitly defined.
In practice, behaviour emerged from a combination of rules, configurations, and manual handling. This made it difficult to predict how conversations would unfold, especially as automation and AI became more involved.
Small changes could have unintended effects, and behaviour varied depending on where conversations started or who handled them. This made scaling support unreliable and hard to control.
Approach
As we explored how conversations were handled, it became clear we weren’t designing flows, but managing side effects of disconnected systems.
This shifted the problem from improving routing rules to defining conversation behaviour as a standalone system.
Instead of extending existing logic, I introduced a central layer where behaviour is defined independently of agents, channels, or interfaces. Conversations can enter from any entry point and move between automation, AI, and human agents within a single model.
As we explored how conversations were handled, it became clear we weren’t designing flows, but managing side effects of disconnected systems.
This shifted the problem from improving routing rules to defining conversation behaviour as a standalone system.
Instead of extending existing logic, I introduced a central layer where behaviour is defined independently of agents, channels, or interfaces. Conversations can enter from any entry point and move between automation, AI, and human agents within a single model.
Decisions
A key challenge was balancing flexibility with usability. Giving teams full control over behaviour could quickly become overwhelming as flows grow.
To address this, I constrained the system through a small set of building blocks, including entry points, branching logic, and escalation paths.
This keeps common interactions simple while allowing advanced behaviour when needed, making the system scalable without becoming unmanageable.
A key challenge was balancing flexibility with usability. Giving teams full control over behaviour could quickly become overwhelming as flows grow.
To address this, I constrained the system through a small set of building blocks, including entry points, branching logic, and escalation paths.
This keeps common interactions simple while allowing advanced behaviour when needed, making the system scalable without becoming unmanageable.
Solution
Chatflow Builder turns conversation logic into explicit, reusable flows.
Teams can define how conversations behave using a small set of components, while a live preview connects flows directly to the end-user experience. This allows teams to understand and validate behaviour before publishing.
By separating behaviour from execution, the system ensures that conversations remain consistent across channels, regardless of where they start or how they are handled.
Chatflow Builder turns conversation logic into explicit, reusable flows.
Teams can define how conversations behave using a small set of components, while a live preview connects flows directly to the end-user experience. This allows teams to understand and validate behaviour before publishing.
By separating behaviour from execution, the system ensures that conversations remain consistent across channels, regardless of where they start or how they are handled.
Outcome
Chatflow Builder introduced a new way to define and control support operations.
Behaviour is now defined in one place and applied consistently across channels, making interactions more predictable and easier to scale.
Teams can expand automation and AI without increasing headcount, while maintaining control over how conversations resolve.
It also established the foundation for Deskpro’s AI-driven messaging, powering both the Messenger and chatbot experiences.
Chatflow Builder introduced a new way to define and control support operations.
Behaviour is now defined in one place and applied consistently across channels, making interactions more predictable and easier to scale.
Teams can expand automation and AI without increasing headcount, while maintaining control over how conversations resolve.
It also established the foundation for Deskpro’s AI-driven messaging, powering both the Messenger and chatbot experiences.
← previous
( 1. )
next →
( 3. )