Loading post…
IntercomExplainer
Principles of Bot Design: Successful Chatbot Design in 2021 [+Tips]
Summary
The post argues that early chatbot attempts have failed because they over-relied on natural language processing and artificial intelligence capabilities that do not yet exist, often making users feel duped. It presents eight opinionated design principles to guide practical conversational UI development, starting with never pretending to be human. Other core principles include keeping conversations incredibly simple and bounded to particular subjects with linear flows, using structured input like buttons and custom keyboards to avoid parsing errors, and respecting the chat medium by showing formatted card previews instead of raw URLs. The post also emphasizes providing a human escape hatch and optimizing for end-user experience rather than support team convenience, with admin views maintaining full conversation context. Finally, it notes that these principles should be strongly held as a decision-making north star but allowed to evolve as technical possibilities improve.
Context
The post was written in 2016 during early chatbot hype when many attempts failed due to overreliance on immature AI, creating a need for practical design principles.
Approach / What changed
The post proposes eight opinionated design principles for chatbot interactions: don’t pretend to be human, keep it incredibly simple, respect the chat medium, optimize for end user, use sparingly, provide an escape hatch, use structured input, and ensure everyone sees the same thing. It provides rationale and examples for each principle.
Takeaways
Early chatbots failed because they relied on natural language processing and AI capabilities that did not yet exist, often bait-and-switching users.
Effective design principles are opinionated and help teams make consistent decisions, unlike truisms which provide no guidance.
Bot interactions should use structured inputs like buttons and custom keyboards to keep responses on track and avoid parsing errors.
FAQ
Why should a bot not pretend to be human?
The post states that pretending to be human makes users feel duped and confused about how the system works. Bot messages should be clearly labeled as non-human, even if the bot has personality.
When is it better to use a human instead of a bot?
According to the post, if a human would provide a better end-user experience, a bot should not be used. Bots should improve what humans are slow at, not replace what humans are good at.
How can structured input improve chatbot interactions?
Structured inputs like buttons and custom keyboards limit the range of possible responses, prevent the need for users to guess correct input, and allow validation before sending, thus avoiding the complications of parsing unpredictable plain text.
Source and freshness
- Source
- intercom.com
- Published
- Added to Yomu
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.