Building a Chatbot Creator That Non-Technical Customers Could Actually Use
Overview
Teams needed a way to create and maintain FAQ modules for Telegram, Slack and WhatsApp without depending on engineering for every content change. The builder had to make modules, datasets, testing and downstream actions understandable in one journey.
Introduction
Turning complex AI into a usable product
Teams needed a way to create and maintain FAQ modules for Telegram, Slack and WhatsApp without depending on engineering for every content change. The builder had to make modules, datasets, testing and downstream actions understandable in one journey.
As product designer on the EasyBits team, I shaped the builder flow and collaborated with Teona and Felix on testing and production validation. A/B experimentation happened later; results were not documented.
Problems & Challenges
Key challenges and the questions we asked to solve them
Building an AI chatbot with EasyBits’ existing tools involved multiple steps, products, and specialized knowledge. Our audience ranged from AI-curious founders to busy marketing and HR teams, so the product could not assume technical expertise.
So, how might we…
…help users create a chatbot in as few steps as possible?
…reduce the need to leave the app for third-party tools?
…show users how their chatbots are performing and allow for fast iteration?
…highlight easybits’ infrastructure without overwhelming users?
Project Goals
What we aimed to achieve for users and the business
Help users build chatbots confidently
Training and deploying AI chatbots was reserved for developers. Our goal was to make this power available to marketers, HR teams, and founders, with no technical friction.
Showcase business infrastructure
The product had to prove the value of easybits’ backend AI platform to investors and early adopters. Every interaction with the app needed to demonstrate capability and potential at scale.
Give users control over tone and data
Users needed more than an AI chatbot; they needed one that felt like their brand. We let them train assistants with their own data and adjust tone, behavior, and messaging without code.
Solution
Turning AI infrastructure into a guided product
From engineering-supported chatbot setup to an observable builder with explicit modules, datasets, testing and actions.
Core product flow
Chatbot Creation Flow
Over five iterations, we refined a guided, step-by-step flow. It allowed users to create functional bots without touching code, while still offering deeper customization where needed.
Customize Data, Variants & Answers
Personalization & Training
Users could train assistants with their own datasets, control brand voice, add question variants, and fine-tune behavior.
What testing changed: usability testing uncovered missing basics—including greetings—which we added before the next iteration.
Empower data-driven decisions
In-App Analytics
We designed a clear dashboard that surfaced metrics like message volume over time. This gave users insight into performance and helped guide chatbot improvements.
Results
What the MVP clarified and why it matters
The work made the setup path observable and testable for non-technical operators. Testing surfaced module confusion, dataset gaps and false-green save states and informed the production validation loop. No adoption or conversion metric is claimed.
Learnings
What this project taught me about builder UX
Builder UX needs explicit feedback at every point where a non-technical operator could lose confidence. Modules, datasets, testing and actions should feel like one understandable journey, with the system making uncertainty visible instead of hiding it.