Sales Buddy

Architecture

Inspiration

Talking to some mom and pop store's the most basic problem to have a better online presence was the difficulty this people had with technology, making it less likely for them to create their own ecommerce and making the best out of it, but one thing they all had in common is the usage of a messaging system (mostly Whatsapp) in their day to day work.

What it does

Our solution would create a basic robot system to help our client to handle their online store all throught a messaging system, so any orders would start a communication from our bot and the flow could automate tasks for our clients employees and even buyers.

Also having a strategy and communication team working directly with our clients to understand their needs and create better automation jobs throght bot conversations with the client, making the possibilities for automation endless.

How we built it

We have basically 3 microservices talking to each other:

MainAPI: Receives users messages and proccess them to find an automatic conversation to start with this user, getting information from the commertools api, registering in the database and calling notifications for this user.
AuthNotify: Handles all users data, authentication and notification beeing by emai, sms, websocket or any other messaging system bots.
CommerctoolsAPI: Handle basic endpoints for querying the store and receives webohooks on orders updates and pass it to the MainAPI.

We are using Contentful to create the conversation's (wizards) logic, so content authors can create new automation process for our clients, and we want to use Algolia NLU (Natural Language Understanding) to check on user's input to figure out what automation process to start, although is still a experimental feature and we couldn't get access to it.

We also have an EC2 instance running our TelegramBot, but we were not able to finish the implementation with our MainAPI due to lack of time.

For email and sms we are using AWS SES and SNS.

Challenges we ran into

The NLU feature of Algolia is very intresting but we took some time to understand we would not have acess to it so the matching for users input still needs a full match with the Wizard title to begin the process. Building an API that could be just used by content authors to create different Wizard like conversations with a user was really challenging with the time we had.

Accomplishments that we're proud of

The notification structure that handles the return messages to the end user depending on wich channel is beeing used and to get a pretty good look of Algolia, Contentful and Commertools with the time we had was really nice.