Getting Started

This guide takes you from your first sign-in to your first training session with Industry Intel data. If you are new to Kaleidoscope, read the Overview first.

1. Sign in to your workspace

2. Invite your team

3. Review the Industry Intel datasets

4. Connect your own data (optional)

5. Set up the SDK

6. Run your first session

1. Sign in to your workspace

Your workspace URL has the form login-<workspace-name>.integrateai.net. The text after login- is your workspace name. Keep track of it, because you need it when you work with Guidewire and partners.

Guidewire creates the first Administrator account for your organization. To activate it:

1. Give your Guidewire contact the full name and work email address of your Administrator.

2. Wait for an invitation email from support@mail.integrateai.net.

3. Follow the instructions in the email to activate your account and sign in.

If your organization requires single sign-on (SSO), ask your Guidewire contact to set it up before you invite other users. Kaleidoscope supports OIDC-based identity providers.

2. Invite your team

Administrators decide who can use the workspace and what they can do. Give each person the role that matches their work:

• Administrator for technical administrators

• Data Custodian for people who only register datasets

• Model Builder for data scientists and actuaries who run sessions

To invite a user:

1. In the side navigation bar, click Settings > Members > Invite member.

2. Enter the user’s details and select their role.

3. Click Invite member. The user receives an email to activate their account.

3. Review the Industry Intel datasets

You don’t register Industry Intel data yourself. Guidewire prepares the Industry Intel datasets for your use case, registers them in the Guidewire workspace, and connects them to your workspace. When a dataset is connected, Guidewire’s connection email arrives with a link to the dataset.

To find a connected dataset:

1. Click Datasets in the left navigation bar.

2. Look for dataset cards with the Partner Dataset label. These are the datasets Guidewire has connected to your workspace.

3. Click a dataset to open it.

4. Select the Dataset schema tab to see its columns and data types.

You can include connected Industry Intel datasets in your training sessions alongside your own data. The schema shows you what your own data must look like to train with them. You use it in the next step.

If you expect a dataset that isn’t on your Datasets page, contact your Guidewire contact.

4. Connect your own data (optional)

If you only want to work with Industry Intel data, skip to Set up the SDK.

To train on your data and the Industry Intel data together, you register your data in your workspace. Your data stays where it is: a task runner processes it in place.

4a. Choose a task runner

If your policy…	Use	Setup
Allows read-only access from Guidewire to an S3 bucket or Snowflake	Hosted task runner	Hosted Task Runner
Requires all computation to run in your own environment	Edge task runner	AWS, Azure, GCP, or on-premises

An edge task runner needs an IT administrator with permission to create roles and policies in your cloud account. Ask your Guidewire contact for the account details the setup steps refer to.

4b. Prepare your data

For training alongside Industry Intel data:

• Use Parquet (preferred) or CSV files under 1 GB.

• Use the same column names as the Industry Intel dataset for every column you want to train on.

• Encode features the same way as the Industry Intel dataset. For example, a given state value must map to the same encoded columns in both datasets.

For the full rules, see Data Requirements and Feature Engineering Requirements. Kaleidoscope’s Transform session can help you prepare the data.

4c. Register your data

Register your dataset on the Datasets page, choosing the task runner you set up:

• AWS

• Azure

• Snowflake

4d. Validate against the Industry Intel schema

Check that your data matches the Industry Intel dataset before you train:

1. Open the Industry Intel dataset and select the Dataset schema tab.

2. Click the download icon at the top right of the schema table.

3. Open your own dataset, select the Dataset schema tab, and upload the schema you downloaded.

4. Click Validate data and wait for the session to finish.

If validation reports problems, see Reviewing validation results.

5. Set up the SDK

Data scientists and actuaries use the Python SDK to run sessions from a notebook.

1. Make sure you have Python 3.10 and pip 22.2 or later.

2. In your workspace, generate an access token and save it somewhere secure. You can’t view it again later.

3. In a virtual environment, install the CLI and the SDK:

pip install integrate-ai
iai sdk install --token <IAI_TOKEN>

For troubleshooting and updates, see Installing the SDK.

6. Run your first session

Guidewire provides notebooks prepared for your use case and datasets. The notebooks are attached to the connected Industry Intel dataset.

To get the notebooks:

1. Click Datasets in the left navigation bar.

2. Click the Industry Intel dataset (labelled Partner Dataset) to open its details page.

3. Select the Attachments tab and download the notebooks.

4. Open the notebooks in the Python environment where you installed the SDK.

A typical first workflow is:

1. Explore the data. Run an EDA session to see summary statistics for each dataset.

2. Train a model. Run an HFL training session using your data and the Industry Intel data. Choose from FFNet, GLM, or GBM models.

3. Compare results. Review the session metrics to see how much the Industry Intel data improves on a model trained on your data alone.

You can follow each session’s progress and logs on the Sessions page in your workspace.

Getting help

Contact your Guidewire contact for help with your workspace, datasets, or notebooks.