Back

AI Chatbot for Better Customer Engagement

About client

Aldi is the common brand of two discount supermarket chains with over 10,000 stores in 20 countries.

Project description

The Aldi client wanted to provide a better tool for communicating with their client and build a closer relations with them. Thus the creation of a chatbot was the logical choice. The bot is able to handle the frequently ask questions, initiate the requests in CRM, help reporting the lost trolleys, reporting complains and many other useful features for Aldi customers.

Project duration: 3 months.

Development team
Architect, Developer, QA and Automation QA.
What we built

The solution includes
Facebook bot used for replacing manual communication for automation between internal stuff.

Integration with SalesForce CRM was used for users’ information gathering and task logging in CRM for further processing. Also it allows requesting chatting with the live agent directly.  Since the user has started to chat with a live agent, bot stopped to work.

The services and Frameworks we used

Heroku – the platform where the project was deployed.
SalesForce is the main project CRM systems. The system gets feedback from clients and automatically divides them into the types and creates appropriate tasks.
Node.js + Express.js were backend API for the bot that works in сhat. It means that all a bot’s dialogs are placed in DialogFlow.
DialogFlow – the service that allows creating a bot that holds a human-like dialog.

Development process

The first steps were to gather the client’s requirements and understand how to build the bot the client needs. During the 1st sprint we built MVP to show a client how bots look in real cases.

After the 1st sprint, we created separate branches – development and separate Heroku environment for testing. This solution allows the client to demonstrate chatbot working process while it is in the development.

The 2nd sprint includes the bot integration with SalesForce system. We completed the complex task of scraping to scrap shops data that located in Australia. Also, there were new additional bot features.

Bot architecture

Technical challenges

Additional message sending. The DialogFlow doesn’t allow for sending one more message after the main request. We found solutions for how to resolve the issue: after preseting the time through the Facebook API the system sends the necessary message to the client.

The result

Technical result: We built the bot with an easy-to-use interface able to change and update texts. For this purpose, a special effort was done for building additional integration between Botkit and DialogFlow. The bot has a lot of notifications and custom messages.

Business result: We satisfied customer demand to build the bot that handles FAQ, initiates the requests in CRM, informs about the lost trolleys, reports complaints etc. The bot improves engagement and builds close relationships with buyers.

Still have questions?

We will give a preliminary assessment within 3 minutes (fill out the form) and are ready to start the project within 4 working days.