Sweet Bot

Intention

Here's a daily scene: a customer browses through a restaurant's website, finds the dishes he wants and makes an order. Later on an employee would process the order, or may rejects when it exceeds the availability.

Life could be easier if there is a bot that assists the whole process - the customers not longer need to access to that particular website, finding the dishes from the page in order to make an order, and the employees can update the supply to avoid oversell, and also get timely updates when an order comes. What's more, a minimal interface removes the needs of investment in website design and host for small business.

Context

With the concept of zero-ui and the API supports for transmitting messages back and forth between endpoints and servers, it is possible to build a bot that handles users inputs and completes all the logics in the backend. Both clients and restaurant employees can benefit from this simplified interface, which is more accessible and has less steps involved.

Process

Bot Personality

Workflow

A plot showing the workflow of handling customers' requests is provided below.

Priority matrix

Here's a list of features designed for the SweetBot

• Menu display: show user what dishes are available.
• Prepare for order: recognize customers' order instructions and parse them into organized list.
• Order confirmation: asking customers for confirmation before settle down the order.
• Make order: update data in database to reflect the effect in order. For example, the available quantity of ordered items should be updated.
• Update menu: accept employees' input to change menu (add/delete/change items)
• Update quantity accept employees' input to update supply for each item.
• Employee confirmation: ask for confirmation before making real update.
• Multiple role support: distinguish between employees and customers.
• Show availability for employees: allow employees to check how many are sold or left for each dishes.
• Database support: use database to store user and dishes data.
• The priority matrix is plotted below.

Product

I use twilio service and slack app api for message delivering. Wit.ai is used in understanding user inputs. 

Reflection

The final product is quite different from what I initially planed to have. I think one of the reasons is that during the process of documenting and analyzing my design, I realized what features are more important in demonstrating the bot's functionality and usability. Those are ones that deserve more effort. 

When integrating the bot with Slack App, there were options in using some builtin interactive interfaces provided by Slack. It was not that useful as it turned out none of the options fit well with my need. I think when designing a product with zero UI, it is important to think about what end points it is designed for. Different platform provides APIs differently, and that could impact on users' experience with the bot.

Improvement 

• Making the response more natural and attractive, such as providing images and/or suggestions
  
• Allow customers to order via slack 
  
• Providing more control for employees on editing categories / orders