Artificial Intelligence Integration: Connecting AI Agent with User

Client

Elinext cooperated with this customer on several projects, dating back as early as 2016. All previous interactions and cooperation were a success, so it came as no surprise that when the customer needed a developer for his new project linked with AI assistant integration, he called Elinext. However, having multiple projects in line simultaneously, he uses multiple vendors for their execution, while never forgetting about us. This time he needed to integrate AI into the website.

Project Description

Our customer has the app and current project developed by the Elinext developer. They asked to implement a demo of a new concept. They wanted to add to the app to understand if it meets investor's expectations. 

The demo should run on a website representing the app's mobile version and communicate with the backend where the AI assistant is running.

The concept was the following:

The backend takes data from databases with a history of user location and access to the user schedule. Sometimes, the backend studies user trends and cashes them to speed up the operation upon request. When the user opens the website, it shows some information about user behaviour. The website had a map with different layers, highlighting our GIS development capabilities.. By default, the map showed the user's schedule. 

The idea was for the app to get the user's trends (based on the places he visits) and adjust his/her calendar, suggesting places to visit when they go to a new city based on his/her visit history.

Challenges

Our customer asked us to implement a demo of a new concept which included the AI assistant integration. Our main challenge was to add a bridge between the AI assistant and the user. For the demo purpose, when the user navigates in a calendar, each day has a different functionality. 

To perform the AI integration in the website, the following functions were to be up for the realisation:

• Search new city: when the user types the city he wants to visit, an AI assistant based on user history and preferences will search for places the user might want to visit in a new city and show them on a map with a description of why this place has been suggested.
• Add a new event to the user schedule: The AI assistant checks the user schedule and finds the most relevant time slot for the new event based on the user's location and gaps between other events. The new schedule is shown on the map.
• Search anything (chat with an assistant): the user can type anything in the chat window (for example "I need a haircut on Wednesday"). The AI assistant will understand the request and will search best places for the request. It is not limited to places, it can also help with location and schedule among other aspects. All AI assistant thoughts and actions stream to the user chat window so the user can understand the process.

Process

We communicated via Slack; Elinext provided the information about the updates.

Our engineer was given a Jupyter Notebook file with some code samples and was given the task of making a backend for their ideas to have a chance to be presented in the form of a demo to demonstrate the AI assistant integration. The process also involved front-end software development to create a seamless user experience.

Solution

The application our customer was building gathers user’s data. Our engineer was given data from a test user. Based on the data of the places he visited during the past two weeks, the user should have gotten AI recommendations on places to visit in a new city. The data was from a Melbourne-based person arranging a trip to Paris, France.

Our developer got the data in JSON format, and Mistral AI with the prompt of AI displaying behavioural trends advice.

Based on the provided location history, AI should suggest places to visit in a new city and the approximate schedule. 

Each place had to be connected with Google Places for the user’s convenience. To get it viewed on a map, we need the coordinates, so that’s why we searched this on API Google Places.

Our task was connecting the agent with the front end of the web application that simulates the mobile app.

While trends are being created, the user gets beautiful animation, and our customer gets the full log on a website of what is going on at the back end.

We used the AI-run agent, which cooked the final suggestions. In the end, users got suggestions of places to visit, and our customers needed that functionality to be demonstrated to investors. This highlighted Elinext’s proficiency in AI solutions, helping our customer present innovative features effectively.

After the demo was completed, to the great success, our customer needed some extra features, like adding the events to the calendar. This was at the second stage of the project, and its demo also was a success.

In the third stage, our customer wanted the “search anything” function to be up for realisation. AI assistant thoughts and actions stream to the user chat window so the user can understand the process, and get to the bottom of it.

Results

Elinext engineer successfully added a bridge between the AI assistant and the user with the help of his impeccable backend development expertise. Our customer successfully introduced the features he wanted to highlight to the investor, and they will be included in the final version of the app. AI assistant integration went smoothly

New features are to be implemented as well in the future. This is an ongoing cooperation, and both parties are mutually satisfied with the cooperation on this and other projects.