What is a Data Search node?

A Data Search node lets your AI agent answer questions using structured data from a file, such as a CSV or Excel spreadsheet.

Think of it like giving your AI agent a simple database. Instead of guessing, the AI searches the data you provide, finds the most relevant rows, and answers based on that information.

Data Search is useful when your business information is stored in tables, spreadsheets, lists, reports, product catalogs, policy records, account data, inventory sheets, pricing tables, or other structured files.

When to use a Data Search node

Use a Data Search node when your information is:

• Stored in CSV or Excel files
• Organized into rows and columns
• Meant to answer specific questions
• Better handled as data instead of long text documents

For example, you can use Data Search to answer questions like:

• “What is the price of this product?”
• “Show me all plans in this category.”
• “What policy belongs to this customer?”
• “What is the status of this order?”
• “Which broker is connected to this policy?”
• “What is the loss ratio for this class?”

What makes Data Search different from Document Search

Use Data Search when the information lives in a spreadsheet or table.

Use Document Search when the information lives in documents, PDFs, articles, manuals, or long text.

A simple rule:

Have rows and columns? Use Data Search.
Have pages of text? Use Document Search.

The Data Search node supports two data source options:

1. Single file
2. Multi-dataset

This means your AI agent can either search one spreadsheet or search across multiple related files.

Single file

Use Single file when all the information your AI needs is inside one CSV or Excel file.

This is the best option when your data is simple and self-contained.

Example:

You upload one product catalog file that includes:

• Product name
• Category
• Price
• Availability
• Description

When a customer asks about a product, the AI searches that one file and responds using the matching row.

Multi-dataset

Use Multi-dataset when your information is spread across two or more files, but those files are connected by a shared value.

This is helpful when different parts of your business data live in separate spreadsheets.

For example:

• One file has policy details
• Another file has claim details
• Both files include a shared policy number

Because both files have a matching value, such as PolicyNo or Policy #, AskHandle can connect the datasets and help the AI answer questions across both files.

This allows your AI agent to answer more complete questions, such as:

• “Show me the bind ratio by broker.”
• “What is the loss ratio by class?”
• “Which policies are connected to this customer?”
• “What claims are connected to this policy?”

How multi-dataset search works

Multi-dataset search connects files using a join key.

A join key is a column that appears in more than one file and contains matching values.

For example:

The column names do not have to be exactly the same, but the values inside those columns should refer to the same thing.

For example, both files may contain the same policy numbers:

When the AI sees the same value in both files, it can understand that the records are related.

How to configure a Data Search node

Step 1: Add or open the Data Search node

In your flow, open the Data Search node.

You can rename the node in the Node name field. This name is only for your internal display, so choose something that helps your team understand what the node is used for.

Example:

• Product Data Search
• Policy Lookup
• Customer Records Search
• Insurance Dataset Search

Step 2: Choose your data source type

In the Data source section, choose how you want the node to search your data.

You can select:

• Single file
• Multi-dataset

Choose Single file if your AI only needs to search one spreadsheet.

Choose Multi-dataset if your AI needs to search two or more files that are connected by a shared value.

Step 3: Configure Single file search

If you choose Single file:

1. Select the file you want the AI to search.
2. Add your AI response instructions.
3. Add search synonyms if needed.
4. Decide whether you want to show the found data to the user.
5. Save your changes.

This setup is best for simple lookups where one file contains all the information needed to answer the question.

Step 4: Configure Multi-dataset search

If you choose Multi-dataset:

1. Click Multi-dataset in the Data source section.
2. In Dataset 1, enter a clear label.
3. Select the first file.
4. Enter the Join key for the first file.
5. In Dataset 2, enter a clear label.
6. Select the second file.
7. Enter the Join key for the second file.
8. Use + Add dataset if you need to connect more files.
9. Save your changes.

The label is simply a friendly name for the dataset. It helps you and your team understand what each file represents.

Example labels:

• Policies
• Claims
• Customers
• Products
• Orders
• Brokers

Step 5: Choose the right join key

The join key is the most important part of multi-dataset search.

Choose a column that:

• Exists in each file you want to connect
• Contains matching values
• Identifies the same record, customer, policy, product, order, or account
• Is clean and consistent across files

Good join key examples:

• Policy number
• Customer ID
• Email address
• Order ID
• Product SKU
• Account number
• Claim number

Avoid using columns that are too general, such as:

• First name
• City
• Category
• Status
• Date

Those values may appear many times and can make it harder for the AI to connect the right records.

Example: connecting two insurance files

Let’s say your business has two files.

The first file contains policy details:

The second file contains claims or performance data:

In this case:

• Dataset 1 join key: PolicyNo
• Dataset 2 join key: Policy #

Even though the column names are slightly different, both columns contain the same kind of value: the policy number.

That lets the AI connect policy details with performance data.

What you configure inside the node

Depending on your setup, the Data Search node may include these options:

Node name

This is the internal name of the node. It helps you organize your flow.

The customer does not need to see this name.

Data source

This is where you choose whether the node should use:

• One file
• Multiple connected datasets

File

This is where you select the CSV or Excel file the AI should search.

If you use Multi-dataset, each dataset needs its own file.

Label

In Multi-dataset mode, each dataset can have a label.

Use a label that describes what is inside the file.

Example:

• Customers
• Orders
• Policies
• Claims

Join key

In Multi-dataset mode, the join key tells AskHandle how the files are connected.

The join key should be a column with shared values between files.

AI Response Style

This tells the AI how to respond after it finds the data.

You can use this to control tone, format, and rules.

Example:

“Answer clearly and briefly. If the data does not contain the answer, say that the information is not available.”

Search Synonyms

Search synonyms help the AI understand different words that mean the same thing.

Example:

• NYC means New York City
• Policy # means Policy Number
• SKU means Product Code

This is optional, but it can improve matching when customers use different wording.

Show Found Data

This option lets users see the exact matched data rows used by the AI.

Turn this on if you want more transparency.

Leave it off if you only want users to see the AI’s final answer.

Best practices

Keep your files clean

Data Search works best when your spreadsheets are easy to read.

Before uploading a file, make sure:

• Column names are clear
• Rows are not duplicated unnecessarily
• Important values are filled in
• IDs and numbers are formatted consistently
• There are no extra title rows above the table

Use clear column names

Good column names help the AI understand your data.

Better examples:

• Customer ID
• Policy Number
• Product SKU
• Order Status
• Renewal Date

Less helpful examples:

• Info
• Data
• Column 1
• Misc

Use stable IDs for multi-dataset search

When connecting multiple files, use a reliable shared value.

The best join keys are usually IDs or numbers that do not change.

For example:

• Customer ID is better than customer name
• Policy number is better than policy type
• Product SKU is better than product category

Test with real customer questions

After setting up the node, test it using questions your customers are likely to ask.

For example:

• “What is the status of order 12345?”
• “Which policy belongs to John Smith?”
• “What is the price of the premium plan?”
• “Show me the claims connected to policy P-1001.”

If the AI cannot find the right answer, check whether the file has clear column names, clean values, and the right join key.

Important to know

• Data Search is designed for structured files like CSV and Excel.
• Single file mode searches one file.
• Multi-dataset mode can search across multiple connected files.
• Multi-dataset mode works best when your files share a clear join key.
• The AI should answer using the data it finds, not by guessing.
• For long text documents, use Document Search instead.