Hexabot
Client
- Hexabot
Project type
- No-Code Conversational AI Platform
What I did
- Product Design
- Research & Audit
- Prototyping
- Design System
Year
- 2025
A no-code / low-code platform for designing, building and deploying conversational AI. Complex dialogue logic is assembled on a visual canvas from pre-made blocks organised into Perception, Understanding and Action — so the people who know what the conversation should say can build it without an AI engineer in the room.
- 01
Businesses need instant support around the clock, but human-only teams do not scale to it and cost more every hour they cover.
- 02
Building a genuinely intelligent bot needed specialised AI and ML talent, which put it out of reach of the teams who actually knew the conversations.
- 03
The people who understood the domain — support leads, product owners — could specify a flow but could not build one, so every change went through a developer queue.
- 04
Existing builders modelled conversation as a decision tree, which breaks the moment a real person says something the tree did not anticipate.
- 01
A visual canvas where a conversation is composed from functional blocks instead of written as code, with the flow legible to anyone who can read the diagram.
- 02
Blocks grouped into three honest categories — Perception, Understanding, Action — so the builder's mental model matches what the system is actually doing at each step.
- 03
A no-code path that does not dead-end: low-code escape hatches where a flow needs real logic, so the platform is not abandoned at the first hard requirement.
- 04
Deployment as part of the same surface, so the distance between changing a flow and having it live is one action rather than a handover.
A support lead building a triage flow
They lay out the intents they already know from their inbox, connect each to an action, and publish — without opening a ticket with engineering.
Handling what the tree did not expect
Understanding blocks catch phrasings the flow never enumerated and route them to the nearest intent, or to a human, instead of failing.
Iterating on a live bot
A flow is edited on the canvas and redeployed from the same screen, so a wording fix reaches customers the same day it is noticed.
Qarta
