Conversational Web

A chatbot that learns by reading your website.

process

user interviews -> needs validation -> paper prototyping -> heuristic review -> medium-fidelity prototyping -> wizard-of-oz testing -> high-fidelity prototyping -> think alouds -> usability testing -> development -> ...

team

Matt Tharp, Jon Linton, Alan deLespinasse, Ira Ladson

duration

2019 / ongoing

Problem Statement

Prospective buyers of a new product or service often have lots of specific questions. It is hard to find these answers on the seller's website, not because the information isn't there but because it is buried deep within the website's labyrinths. Providing access to knowledgeable sales people via chat 24x7 is expensive, and many sellers consequently lose out on prospective buyers.

Outcome

Conversational Web is "A.I. in chatbot clothing." It lives at the bottom of the page, but instead sending your questions to a live customer service team, it answers (most of) them right away. When it fails, it elicits human help. Using contextual clues based on the website's DOM, Conversational Web also surfaces relevant prompts to engage visitors as they explore a company's website. The beta version of the product will launch in January 2020. More info here.

Selected Assets

A selection of assets from the prototyping process are presented here. As stated above, this is an ongoing project.

[Figure 1] A collection of paper prototypes from the early stages of the project.


[Figure 2] A low-fidelity wireframe showing the form factor of the chat window.


[Figure 3] A wireframe showing the chat window in the middle of a conversation.


[Figure 4] A high-fidelity prototype showing the chatbot attempting to engage a visitor with a proactive question. The question fades when the timer runs out after 10 seconds.


[Figure 5] A high-fidelity prototype showing the chatbot attempting to engage a visitor with subtle calls-to-action (CTAs) as they scroll down a webpage. The CTAs are contextual, i.e. related to the content that the user sees on the webpage.


[Figure 6] A high-fidelity prototype showing a conversation between the chatbot and the visitor.