iPaaS meets Artificial Intelligence: A Case Study

An iPaaS as the name suggests is a set of tools and services given to the user as a platform that can help in developing integration solutions. APPSeCONNECT being an iPaaS is capable of connecting any business application or data sources to make sure the data seamlessly transfers from one application to another. Now sometimes the data we transfer from one application to another is not straightforward and requires additional plumbing to get more insights about the data. Traditionally people used to develop services that keep manual rules to identify the data insights and generate raw output, but things have been changed recently.

With the proliferation of data, people are building services that can specifically give you intelligence on the data, some are even publicly available and can be consumed readily without any issues. In this article, we will consume a readymade AI-based Cognitive API from Microsoft to get insights about the tickets, such that one can prioritize the tickets coming from real customers in Salesforce and get some readymade insights without reading the whole ticket entries.

Here is what you can achieve through this article :

1. Get the tone of the Ticket and determine the urgency. For example, you can check the negativity of the ticket text such that if the customer is very much complaining, you need to prioritize the ticket.

2. You can assign the ticket to the correct Agent based on the ticket content.

3. You can identify the main points of the ticket without reading the whole text.

There are many more things you can do using the services, but we are restricting our article to only these.

In this article, we will show you how you can use our REST connector to connect a Cognitive Service to get business intelligence of the tickets. The purpose is to give knowledge on how Generic Rest Adapter could be implemented easily for integration between chosen applications. Here the chosen applications are – Salesforce CRM and Microsoft Cognitive framework using ProcessFlow.

If you have multiple systems running in your business, connect all of them under one single platform to automate the business process and improve your productivity and efficiency!

Introduction: iPaaS meets Artificial Intelligence

Integration is one of the most common needs for driving any business successfully. We invest a lot in finding the best tools that can help us to seamlessly sync up information between applications. APPSeCONNECT is one such integration platform, which provides a simplified approach to capture data, connect disparate applications and perform actions.

In this article, we will discuss some of the important facts that you must consider while working with APPSeCONNECT, as the integration tool to sync data between applications. We will go ahead and configure the APPSeCONNECT platform to connect Salesforce with my File System.

From the article, we will try to cover,

1. How to connect an external application with the APPSeCONNECT integration platform when the application is natively supported, in our case it is Salesforce and Cognitive Services.

2. And obviously, we will give you some pointers which help you to get most of the platform.

Background

If you are targeting the emerging market of citizen integrators, the need for tools to empower them to integrate almost any application is a must-have for your company. As citizen integrators have their domain expertise, they are always on the lookout for the solution that equips them with self-service data integration with better data visualization to make their work much easier and more functional without being much technical.

Business scenario

Detecting Customer Sentiment from Salesforce using Cognitive Services

Customer relationship management is critical to any business. Salesforce, being a pioneer in CRM solutions, has a lot of functions to take care of any customer in a business. Salesforce ticketing system is the grievance cell of a customer where they can go and provide an issue or feedback. The agent is having a responsibility to work on an incoming ticket and respond quickly.

Traditionally, in a business, the priority and severity of tickets are used to identify a ticket type and time criticality, but in most cases, customers do not feed those values. Thus there is always a need by the agents to identify the priority of a ticket to serve their customers better.

A bad sentiment expressed by a customer is a red flag. If agents could not serve that customer quickly, the customer might eventually churn. Hence it is critical to identify the sentiment of a customer to serve the customer faster and better. With the proliferation of a large number of online services, we thought to use Cognitive Services to identify the sentiment of the customer such that we can pick the right ticket from the list.

In this article, we consider three nodes which are Salesforce, Microsoft Azure Cognitive Services and File System. A new ticket is fetched from Salesforce will be passed to the Cognitive APIs to detect its sentiment and then the output will be stored into a file in the local file system. The system will allow the user to understand the sentiment score of the ticket and based on this one can explore infinite other possibilities to automate certain features in Salesforce.

We will give you pointers too, on how to effectively utilize your skillsets to automate certain other behaviors in this module.

Prerequisites

Before starting the integration using APPSeCONNECT, you must understand that you require a few pre-requisites for establishing integration between these applications such that you can effectively connect to any REST APIs. As mentioned in the beginning, example of Salesforce application integration with File System has been taken in this article to explore the features that APPSeCONNECT offers to make connection simpler. So, from both the endpoints there are certain pre-conditions that are required to be fulfilled. Let’s have a look:

You must have a valid Salesforce developer account.
You must have a valid URL of Microsoft’s Cognitive Services.
You must have a valid path for read and write using the File System.
You must have a registered account with APPSeCONNECT.

Acquiring Salesforce Credentials

Follow the steps to get the Salesforce Credentials:

1. To access Salesforce API, you need to register to a Salesforce developer account. Click here to register to the Salesforce developer account.

2. Provide account details and create the developer account.

3. Once you create an account, you will get the activation email. You can click on the link provided to activate the account.

4. Now use the username and password to log in to the account.

5. Go to this settings icon and click on the Setup option.

6. Next go to the Apps section on the left side and then App Manager.

7. After that go to New Connected App at the right top And fill up all the required field and save it.

Note The generated client id, client secret and redirect URI will be used later during credential configuration while using APPSeCONNECT.

8. Keep note of the Salesforce URL, as this information will be required for configuring Salesforce in APPSeCONNECT.

Acquiring Azure Cognitive Service Credentials

After getting the salesforce credentials, it is important to get the credentials of the Cognitive framework too. We are using the Text Analytics service for time being. You can get a full list of APIs here.

1. Open Azure Portal.

2. Create a new Service, choose Cognitive Service.

3. Provide the required details, and get the keys from the account.

4. Once the deployment is done, you can open the service and get the Keys. These keys you need to keep handy as these will be used to call the cognitive services.

Acquiring File System Folder

Now, this is very easy to have. Just create a folder in your drive and we will use this to store the results. You can directly update in salesforce too, but for this article, we will keep the result in a file.

Getting Started with APPSeCONNECT

Finally, when you have all the credentials, you need an APPSeCONNECT account to do this transformation. If you have an account already go to portal.appseconnect.com and log in.

If you don’t have an account, feel free to reach us for a free trial.

Implementation

The implementation of the process through APPSeCONNECT is easy and has a very little learning curve. We will take step by step approach to do such an operation.

1. Open APPSeCONNECT portal.

2. Go to Designer -> ProcessFlow.

3. Click Create to create a new ProcessFlow.

4. We name it “Salesforce – Cognitive Sentiment analysis”. Feel free to create your own.

5. Drag a Start Node, Configure it for Pooling.

6. Drag a REST connector from Flow Connect of the toolbar.

7. Once you drag a REST connector in the canvas, you will be asked to provide the name of the App. Name it anything that suits your requirement. For us, we name it “Sentiment Analysis”.

8. Once the App is created, you can import the schemas. To Import, just go to the service documentation site, get an output file and save it in the local machine.

9. Finally, go to “Manage Schema” section and import the schema.

10. We name it as documents. Generally, the Entity name does not matter much, but it is good to choose it wisely.

Once the mapping is done for rendered entities, please click Save to continue.

11. Once you have added the schema, you need to add Action.

12. Adding actions will enable you to provide endpoint information, action and error filter details that are required to fetch or post data in any of the respective apps. Click on the Action node and add an action.

13. The action filter for the action requires an OCP-Apim-Subscription-Key header that connects to the Cognitive Service APIs. This is the key you got from the Azure Portal.

14. Now configure the salesforce Action filter too as mentioned in this document to get the ticket information. This will be required to Get the data from “Salesforce”.

15. Once salesforce is configured, drag a File System node, and do the setup. Configuring the file system app is also very easy, you are about to create a schema to push the data and add an action.

16. Once all are done, drag mapper nodes to link between these apps.

17. The result looks like this. The process is, ticket data is fetched from Salesforce, then processed into a Sentiment analysis node which sends the data to the Cognitive Service. The output of the data is stored into the local file system and also synched back to the salesforce again to update the ticket metadata.

You can now deploy and execute to get the data synced.

Execution Results

Now after we execute the ProcessFlow, the Salesforce Ticket is being fetched and then send through the Sentiment Analysis Service to find its sentiment. The service gives us a result on the overall sentiment of the ticket, as well as the sentiment per sentence. The confidence Scores will indicate whether the ticket is Positive, Neutral or Negative.

Now for this particular execution, the sentiment is treated as negative. Value 1 in the negative element under ConfidenceScores indicates that the negativity is confirmed. You can see the Text which we have passed to the AI service is “Not Good” which correctly resulted in Negative.

Similarly, if we show you another record,

Here the text sent is “Beautiful” which is indeed a positive word and the result is identified properly.

To understand the positivity, we have taken a hypothetical logic based on our assumption.

1. Any value from 0.5 to 1 is treated as 1 in confidence, any value from 0 to 0.4 is treated as 0 in confidence. (We took this for simplicity).

2. If Confidence is Positive, we mark the Ticket as Low.

3. If Confidence is Neutral, we mark the Ticket as Normal.

4. If Confidence is Negative, we mark the Ticket as High.

5. The actual sentiment is analyzed based on the first occurrence of confidence from bottom to up. That means if we get both Neutral and Negative, we take the Negative as actual value. Similarly, if we get both Positive and Neutral, we take it as Neutral.

This way we have simplified the values correctly and marked the ticket priority. But as the possibilities are infinite here, we can customize this logic anyway to identify the actual sentiment more effectively.

References

To know more about how to work with the nodes here are few links which can help you.

Conclusion

In modern days with the proliferation of AI and machine learning techniques, it becomes easy to get into this space to develop innovative solutions. Customer relationships could be improved manifolds just by doing simple things right. From this article, we have shared a scenario that could really help your customer getting the right services at the correct time and also made you understand how to connect to a REST-based API directly using APPSeCONNECT and also the file system. Feel free to ask your questions, I would be happy to answer.

Thanks.