Install and configure

The page provides instructions on installing and configuring the Kayla AI Agent for coordinating the perpetual Know-Your-Customer (pKYC) process.

Check requirements

WorkFusion offers the following options to install and run Kayla:

• On your premise
• In your cloud
• In the private cloud hosted by WorkFusion

System requirements

The server requirements for Kayla, the pKYC Coordinator, are the same as for Work.AI (IA Cloud Enterprise) platform. For details, refer to the Hardware and OS article.

Hardware

• No RPA bots required

• Proxy for high-availability installations

• No server for analytics required

Software

• IA Cloud Enterprise v10.2.7 or higher

• No machine learning (ML) or analytics component used

Install Kayla

To install Kayla, obtain an Asset Bundle (zip archive) with the required release version from the appropriate S3 bucket. Then, upload it manually in the Control Tower interface in accordance with the guide.

Once the installation is complete, you can proceed to configure Kayla's settings.

note

You can configure multiple variations, or sets of parameters, for a single AI Agent. For example, different departments can set up and manage different requirements or input methods per variation. For details, refer to Manage AI Agent variations.

Configure Kayla

There are two possible ways to start the configuration flow for Kayla:

• If you install Kayla without a preset configuration, wait for the installation to complete. Once you are redirected to the AI Agent version details page, open the Actions menu and click Settings.

  

• If you install Kayla as an Asset Bundle with a preset configuration, wait for the installation to complete, and click the Show digital worker button. Then, find the installed variation in the Digital Workers list and click Prepare.

  

After you complete any of the flows described above, proceed to the configuration:

1. In Advanced > Data Stores, set up a proxy server.

2. On the AI Agent configuration page, perform the following steps:

   1. Configure the input settings.

   2. Specify data enrichment sources.

   3. Configure third-party data sources.

   4. Set the review process rules.

   5. Configure a human-in-the-loop step (optional).

   6. Define the way to receive Kayla's output.

   note

   When you configure your AI Agent variation for the first time, you must complete the steps one by one. Once you are in the last step, you can go back and make changes to any previous steps. If, for some reason, you cannot finish the configuration, click Save & continue later. This way, the entries you made are saved, and you can go back and complete the process later.

Proxy server

To configure a proxy server, do as follows:

1. Go to Advanced > Data Stores and select the uc_DW_KAYLA_KYC_technical_configuration Data Store.

   

2. Set up the following fields:

   Name	Sample value	Description
   proxy.host	proxy.customerdomain.com	Enter your proxy host.
   proxy.port	1234	Enter your proxy port.
   proxy.credentialsAlias	kayla_proxy_credentials	Specify the alias of the proxy credentials in Secrets Vault. Configure only if the proxy server requires authentication.

   note

   If the proxy server requires authentication, set up proxy credentials in Secrets Vault. For detailed instructions, refer to Set up secret entries in Secrets Vault.

Input

In the Input step, you set the input source that Kayla monitors for updates and the rules for how Kayla does that. The input source can be a MinIO bucket, an email box, or both. In any case, the expected input format is a CSV file corresponding to specific requirements.

To configure the input source and rules, follow the steps below:

1. Specify the header names to match the three required columns from the input CSV file:

   • Unique Identifier
   • Entity Name
   • Entity Type

   

2. Configure the monitoring settings:

   • Specify how frequently, in hours or minutes, Kayla polls the input source for new files.

   • If needed, select the Stop monitoring after some iterations checkbox. In the displayed Max loops field, specify the number of iterations.

   • Optionally, select the Stop monitoring after some duration checkbox. In the displayed Max Monitoring Duration and Time Unit fields, set the period duration.

   

3. Choose one or both input methods and specify required settings:

   • Shared File Location: set the input S3 bucket and file path in MinIO (S3).

     

   • Microsoft Graph API Email: specify such SMTP settings as the monitored email address, email folder, and Tenant ID.

     

     SMTP credentials are stored in Secrets Vault, so you need to choose a Secrets Vault alias from the drop-down list in the Microsoft Graph API Secret Vault Alias field.

     To create a new alias, click the Create New Secret Vault Entry button, set the alias up, and save it. For detailed instructions, see the Set up secret entries in Secrets Vault.

     

4. Click Next. The step settings are saved, and you are navigated to the Enrichment Sources step.

Enrichment Sources

The step is designed to supplement the input data without direct integration with a third-party tool. In this case, an enrichment source is a CSV file uploaded to an S3 (MinIO) bucket. You can set up as many enrichment sources as you need.

To configure an enrichment source, follow the steps below:

1. On the Enrichment Sources step, click Add Another.

   

2. Specify the location of the file with enrichment data in S3 (MinIO).

   

3. Map the unique identifier in the input file to the one in the enrichment file so that the AI Agent can link the information correctly.

The specified enrichment source becomes visible in the Enrichment Data table from where you can edit or delete it. To add more enrichment sources, click Add Another and repeat the steps above.

Data Sources

Kayla can retrieve information from a variety of data sources out of the box:

• Companies House

• Department of Labor (DoL) EFAST 5500

• NASDAQ

• NYSE

• Sayari Graph

• SEC Edgar

• TwelveData

On the Data Sources page, choose one or more sources in the appropriate box and specify their settings.

The figure below illustrates how the step looks with all available sources chosen:

For the TwelveData, Companies House, and Sayari Graph data sources, you need to choose a Secrets Vault alias from the API Key Alias drop-down list. To create an alias, click the Create New Secret Vault Entry button, set the alias up, and save it. For detailed instructions, see Set up secret entries in Secrets Vault.

Select the Enrich data box and, in the Enrichment Source field, enter the path to the file with enrichment data.

Once you configured all sources as required, click Next to proceed to the Requirements step.

Requirements

When Kayla processes input, it reviews each row in a file and verifies if the KYC data has changed. All data from the input file is passed through to the periodic review form. However, only the input file fields configured on the Requirements page are reviewed and flagged for manual review.

To configure a requirement, follow the steps below:

1. On the Requirements step, click Add Another.

   

2. Specify the rules that Kayla should follow when reviewing KYC data for an entity:

   

   In the table below, you can find more details about the rules you need to set:

   Field	Description	Values	Required
   Entity Type	The type of the screened entity	Pension Plan, Mutual Fund	Yes
   Source	The external source from where input data is retrieved	As configured in the Data Sources step	Yes
   Requirement Reference ID	A key or reference to the KYC requirement matrix	Free-text field	No
   Review Method	The method used to complete the review	Exact Match: the Field Name value is a direct check against a value in the API response from the configured source. Information Extraction Model: the Field Name value is validated using an information extraction model. Manual Review: the Field Name value is not reviewed. Used if an information extraction model or API is unavailable or a data point requires additional scrutiny. Name Match: works similarly to Exact Match, except that it does not consider punctuation, extra spaces, or other special characters that are inadvertently entered in a name. This ensures that minor discrepancies (- vs. : or extra spaces) do not get flagged for manual review. Verify Status: used to verify the Field Name status. For example, whether the entity is regulated and found in a regulator's database and whether the status is a specific value. Verify Address rules model: used to verify complete normalized addresses. It assumes the data is in a specific format, depending on how many address lines are provided, for example: For six address lines: street_line1, street_line2, city, state, country, zipcode. For five address lines: street_line1, city, state, country, zipcode. For four address lines: street_line1, city, state, zipcode. The rule model takes a risk-based approach and accepts the following discrepancies: A missing country in one address is ignored if the rest of the address matches (including zipcode). This is due to some sources that do not include a country in addresses. Country and state ISO codes and all official names are compared when matching. Common aliases, for example, China vs. Republic of China, are also accepted. For example, if address 1 has "USA" and address 2 has "United States of America, (The)", it is considered a match. State alpha codes, ISO codes, and full ISO names are compared when matching. Verify Address ML model: used to verify two addresses. Each address is either normalized (unstructured) or structured.	Yes
   Model	The ML model to be used	Available pre-trained models	Yes, if Review method is set to Information Extraction Model.
   Field Name	The field name to be reviewed in the input file	Depends on the input file	Yes
   Source Field Name	The field from the specified source to be compared to the Field Name from the input file	Varies, depending on the source	Yes, if Review method is set to Exact Match or Information Extraction Model.

3. Click Save to save the rules for the entity.

4. Repeat Steps 1 to 3 to add as many entities as you need to review. The resulting page looks similar to this:

   

5. Click Next to proceed to the Output step.

Human-in-the-Loop

The step is optional. It allows analysts and quality assurance specialists to review Kayla's results in Workspace and address any exceptions prior to the generation of the final periodic review report.

To configure the step, follow the instruction below:

1. Choose an option to define when manual review should be enabled:

   • Enable in all Cases creates an assignment in Workspace for every entity, regardless of whether Kayla completed requirements or not.

   • Disable in all Cases disables manual reviews completely. In this case, the Automatically route users to the next task in the queue?, Do you want to pre-assign work to users, and Case Details settings are unavailable.

   • Disable if all Requirements are Successfully Completed conditionally creates a manual review assignment when there is some investigated requirement that Kayla cannot close with confidence or for which Kayla does not have relevant information.

2. Specify the task routing mechanism. Select Yes for Automatically route users to the next task in the queue? to route users to the next task in the queue that is not actively being worked on, regardless of who it is assigned to.

3. To pre-assign the manual review assignment to a specified user in Workspace, select Yes for Do you want to pre-assign work to users. In the Username field, specify the user to whom you want to pre-assign the task.

4. In the Case details section, set up the required and optional fields to be used and displayed in the manual review assignment in Workspace:

   • Case ID
   • Case Reason
   • Entity ID
   • Entity Name
   • Entity Type
   • Current Risk Rating
   • Previous Risk Rating
   • Jurisdiction
   • Minio UI Base URL

   To include the fields, enter the column headers from the input file for each item.

For instructions on working with human-in-the-loop assignments, see the Review results guide.

Output

In the Output step, you set where Kayla should send its output:

• A periodic review status report
• An entity periodic review report
• A batch summary report

For details about the reports, refer to Input and output data.

To complete the output setup for Kayla, configure the email address where the summary report is to be sent:

• If Email is configured as an input method (see the Input step), reuse the same setup as for the Input step or specify a different email to receive reports.

• If Shared File Location is used as the only input method, set the email for receiving Kayla's reports on the Output page.

  

  When setting the output email, choose a Secrets Vault alias from the drop-down list. If you don't have one, click the Create New Secret Vault Entry, set the alias up, and save it. For detailed instructions, see Set up secret entries in Secrets Vault.

  

Once you have set up the above, click the Finish button to save the configuration.

Update configuration

To update Kayla's configuration, follow the steps below:

1. Go to the Digital Workers tab.

2. Choose the Kayla's version you are working with.

3. On the version details page, open the Actions menu and select Settings.

   

4. On the displayed configuration screen, make changes in each step as required and click Finish to save the configuration.