This project is developed to showcase how Amazon comprehend can be used to classify the incoming email and respond back to the customer automatically.

Supporting customer via email channel is one of the big task for every organization sinch it involves human resources, space, devices,and other technologies. Also, this is very expensive process and not addressing customer queries leads to huge customer churnout and impacting the revenue and reputation of the organization.

Many of a times, customer asks the basic questions and it is unnecessary that customer care rep to respond to all the emails unless the request is critical or sensitive. Amazon Comprehend is managed AI service which helps to extract insights from the content of documents. In this project, we used Amazon Workmail to receive the customer email and that triggers the lambda function to pass the content to Amazon Comprehend to classify content and respond back to the customer via email with pre-defined answer for the known requests.

The solution comprised of two cdk stacks.

• comprehend-custom-classifier-dev-notebook-stack : Creates the Amazon sagemaker jupyter notebook instance pre-loaded with .ipynb notebook and creates IAM role required for executing comprehend custom classification training, deployment, and S3 data access.
• workmail-organization-domain-user-dev-stack : Creates the Amazon workmail with domain, user, inbox access.

• AWS CLI >= 2.2.25 (Please follow Installing or updating the latest version of the AWS CLI guide to install/upgrade AWS cli)
• AWS CDK command line utility (1.120.0) (Please follow Getting started with the AWS CDK guide to install/upgrade cdk.)
• Python>=3.7

Note: You can deploy these stacks in us-east-1(N.Virginia) or us-west-2(Oregon) or eu-west-1(Ireland) (Amazon workmail available only in these 3 regions)

1. Clone the repository.

git clone [email protected]:aws-samples/email-response-automation-comprehend.git

2. This project is set up like a standard Python project. To create the virtualenv it assumes that there is a python3 (or python for Windows) executable in your path with access to the venv package. create the virtualenv using following command.

$ python3 -m venv .venv 

3. Use the following step to activate your virtualenv.

$ source .venv/bin/activate

If you are a Windows platform, you would activate the virtualenv like this:

% .venv\Scripts\activate.bat

Once the virtualenv is activated, you can install the required dependencies.

$ pip install -r requirements.txt

4. Deploying the solution :

Deploying new Sagemaker Notebook Instance with IAM Roles and pre-loaded .ipynb notebook : Execute following command by passing optional paramaters

cdk deploy comprehend-custom-classifier-dev-notebook-stack  --parameters NotebookName=<Notebook Name> --parameters NotebookConfigName=<Notebook Config Name> --parameters RoleName=<SageMaker execution role name>

Arguments to the stack creation :

• NotbookName :(optional) Name of the notbook instance. If not entered, default name 'notebook-instance-comprehend-training' will be used.
• NotebookConfigName :(optional) Name of the Notebook config. If not entered, default config name 'notebook-lifecycle-load-notebook' will be used.
• RoleName : (Optional) Name of the Amazon Sagmaker Execution role name. If not entered, default role name 'sagemaker-notebook-execution-role' will be used.

Deploying new Workmail domain, user, user registration and inbox: Execute following command by passing optional paramaters

cdk deploy workmail-organization-domain-user-dev-stack --parameters OrganizationName=<Organization Name> --parameters UserName=<Support Username> --parameters PassWord=<Password>

Arguments to the stack creation :

• OrganizationName :(Required) Name of the workmail organization. If not entered, default name 'my-sample-workmail-org' will be used. Domain also will be created using this organization alias name. So make sure to use unique alias to avoid errors due to duplicate domain names.
• UserName :(optional) Name of the your organization support user alias. If not entered, default user name 'support' will be used.
• PassWord : (Optional) Password for the UserName. If not entered, default password 'Welcome@123' will be used.

Note : Please note that these both deployments approximately 20 to 25 minutes

After the stacks are succefully deployed (You can see if there is an error as the cdk output otherwise it says stacks creation succeful.), please open the .ipynb notebook from Sagemaker notebook instance and execute all the scripts in the notebook in sequence.

1. Go to AWS console and select the service 'Amazon Sagemaker'. Maker sure you are in us-east-1 region
2. Select the Notebook menu and choose Notbook Instances
3. Click 'Open Jupyter' from the Action menu
4. You can see the notebook 'notebook-instance-comprehend-training' and click to open the notbook
5. Run the script as per the instruction given in the notebook. Classification endpoint and Entity Detection endpoint will be created. Please use these ARNs as the context variables for the next stack deployment.

Below CDK Deployment will create AWS Lambda fucntions, Amazon SES and SNS notifications: Execute the following command by passing the context variables

cdk deploy email-reponse-automation-workflow-stack --parameters emailClassificationEndpointArn=<email classification endpoint ARN from sagemaker notebook>  --parameters humanWorkflowEmail=<Workflow Email> --parameters supportEmail=<support email id created part of the workmail-organization-domain-user-dev-stack> --parameters emailEntityRecognitionEndpointArn=<Enitity Detection endpoint ARN from sagemaker notbook>

Arguments to the stack creation :

• emailClassificationEndpointArn (required) : email classification endpoint ARN from sagemaker notebook
• humanWorkflowEmail (required) : email id to receive the SNS notification if customer email content does not match with any classifcation. The email id will subscribe from SNS topic and SNS will publish unclassified email to the topic.
• supportEmail (required) : Email id created part of the workmail org and user creation. This email id will receive email from the customer and invoke the lambda function
• emailEntityRecognitionEndpointArn (required) : email entity recognition endpoint ARN from sagemaker notebook

1. Click on ‘Organization Settings’ from the menu and in General section you will see the web-application link for your email.

2. Click on that web-application URL to login and enter the username and password and Sign In.

3. You will see your Email inbox web page.

4. Test this new email box by sending email from your personal email id.

5. Setting up Inbound Rules:

• Click on ‘Organization Settings’ and choose the ‘Inbound Rules’ menu
• Click the button ‘Create rule’
• Enter the Rule Name and Select the action ‘Run Lambda’ and choose ‘Custom Lambda function’ and select the lambda function generated in the previous stack deployment.
• This Lambda function will be invoked upon receiving the email on this domain
• Enter * in both Sender domain and destination domain

Currently 3 type of custom email classifications will be used for automated response to customer email.

1. MONEYTRANSFER - Customer email contains query about the money transfer
2. PASSRESET - Customer email contains query related to login, account locked or password request
3. PROMOCODE - Customer email contains query related to discount or promo available for the money transfer If the customer email does not belong to any of the above classifcation, customer email will be sent to SNS topic and whoever subscribe the topic will receive the message. In our testing we subscribe with email we passed with human_workflow_email parameter(ex: your presonal email) to verify the mail has been moved to SNS topic.

• Step1: send the email to the customer support email box (ex: [email protected]) from your personal email to query about the money transfer status.
• Example email content:

Subject: 'Money Transfer Status'  Body: 'Hi Team , Can you please check the status of money tranfer I sent to India? This is my trx id : MTN0000123 Regards CustomerX'

• Step2: Check your workmail inbox and you can see the email sent from your personal email.

• Step3: You can see the automatic email response back to you personal email with your money transfer status with same subject.

Dear Customer, Hope you are looking for the status of your money transfer. Status of your MTN0000123 is INPROGRESS. If this is not right information you are looking for, please reply back to this email with little more information

• Step4: Try sending email with your own subject and body related to above three classications. You will see the different email response based classification results.

• Step5: Try sending email with your own subject and busy not related to above classifications.

Ex: Hello, can you help me to solve my problem?. You will see your email has been moved to SNS topic and you can confirm that by see the email content in your human workflow email you given during the stack creation.

To avoid incurring ongoing costs, delete the resources you created as part of this solution bye executing following commands in order.

1. cdk destroy email-reponse-automation-workflow-stack

2. cdk deploy comprehend-custom-classifier-dev-notebook-stack

3. cdk deploy workmail-organization-domain-user-dev-stack

See CONTRIBUTING for more information.

This library is licensed under the MIT-0 License. See the LICENSE file.

Enjoy!