Should I use AI Answer, Document Search, or Data Search?

Use AI Answer, Document Search, and Data Search for different types of information.

A simple way to think about them is:

• Use AI Answer for conversational guidance, instructions, and behavior
• Use Document Search for written knowledge such as policies, guides, FAQs, and manuals
• Use Data Search for structured information such as spreadsheets, records, prices, and inventory

Choosing the right source makes your AI agent easier to control, test, and maintain.

Quick guide

Use AI Answer for conversational guidance

Use AI Answer when the AI agent needs to:

• Respond conversationally
• Follow instructions
• Explain services
• Guide the customer
• Use enabled skills
• Apply business rules
• Handle follow-up questions naturally

AI Answer is best when the information is not mainly a large document library or structured dataset.

Example

A customer asks:

“Which of your services is best for a first-time customer?”

If the response depends on your service guidance, business rules, and conversational explanation, AI Answer is usually the right place.

What should go into AI Answer?

AI Answer is useful for:

• Role and behavior
• Tone
• Business rules
• Short explanations
• Small amounts of stable knowledge
• Guidance on what the AI agent should do next
• Instructions for when to use skills

For example:

“Explain our three service tiers briefly and recommend the most appropriate one based on the customer's needs.”

That is a good AI Answer instruction.

Do not use AI Answer as a large document library

If you have many pages of policies, manuals, FAQs, or written knowledge, use Document Search instead of placing everything into AI Answer.

Large knowledge libraries are easier to manage as dedicated documents.

This also helps keep the AI Answer instructions focused on behavior rather than content storage.

Use Document Search for written knowledge

Use Document Search when the answer should come from written content you manage.

Common examples include:

• Policies
• Manuals
• FAQs
• Internal documentation
• Help articles
• Service guides
• Product documentation
• Employee procedures
• PDF documents
• Written reference material

Document Search is designed for information that reads like a document.

Example: cancellation policy

A customer asks:

“What happens if I cancel less than 24 hours before my appointment?”

If the answer is stored inside your cancellation-policy document, use Document Search.

The AI agent can search the relevant content and respond based on that source.

Example: product manual

A customer asks:

“How do I reset this device?”

If the instructions are stored in a manual or support guide, Document Search is the appropriate source.

Use Data Search for structured information

Use Data Search when information is organized into rows and columns.

Common examples include:

• Product catalogs
• Prices
• Inventory
• Vehicle records
• Location data
• Customer records
• Insurance factors
• Service rates
• Order information
• Reference tables
• Structured lookup data

Data Search is best when the AI agent needs to retrieve specific values or records.

Example: product price

A customer asks:

“How much does model A25 cost?”

If your product data is stored in a spreadsheet with columns such as:

• Product name
• SKU
• Price
• Availability

use Data Search.

Example: inventory

A customer asks:

“Do you have the black version in stock?”

If color and availability are stored in structured inventory data, Data Search is the appropriate source.

Example: insurance factor

A calculation may require a factor based on engine power.

If each engine-power value is mapped to a specific factor in a table, Data Search can retrieve the correct factor.

The result can then be used in a calculation.

Documents vs. structured data

The easiest distinction is often how the source information is organized.

Use Document Search when the information looks like this:

Cancellation Policy
Customers may cancel up to 24 hours before the appointment without a fee. Cancellations received later may be subject to...

This is written knowledge.

Use Data Search when the information looks like this:

This is structured data.

What if a spreadsheet contains long text?

A spreadsheet is not automatically a Data Search source just because it is an Excel or CSV file.

Think about how the information needs to be used.

If each row represents a record and the AI agent needs to find exact values, Data Search is usually appropriate.

If the spreadsheet is mostly long paragraphs of written knowledge, a document-style source may be more suitable.

Choose based on the information structure and task, not only the file extension.

What if a document contains tables?

A document can still be appropriate for Document Search if the main purpose is written knowledge.

For example, a policy PDF may contain:

• Explanatory text
• Rules
• A small fee table
• Notes and exceptions

If customers mainly ask questions about the policy as a whole, Document Search may still be the better choice.

If the AI agent needs to perform exact repeated lookups against a large table, structured Data Search is usually more appropriate.

Use AI Answer and Document Search together

AI Answer and Document Search can work together.

For example:

AI Answer

• Controls tone
• Defines business rules
• Handles the conversation

Document Search

• Provides detailed policy information

A customer asks:

“Can I cancel my booking?”

Document Search finds the relevant policy.

AI Answer can present the result according to your communication rules.

Use AI Answer and Data Search together

AI Answer can also use structured information as part of a broader conversation.

For example:

Data Search

• Finds the correct product
• Retrieves the price
• Checks availability

AI Answer

• Explains the option
• Answers follow-up questions
• Guides the customer to the next step

This is useful for product discovery, pricing, inventory, and other structured use cases.

Use Document Search and Data Search together

Some AI agents need both written knowledge and structured records.

For example, an insurance AI agent might use:

Document Search

• Coverage descriptions
• Policy explanations
• Claims guidance

Data Search

• Rate tables
• Vehicle factors
• Location factors
• Structured pricing data

The customer's question determines which source is appropriate.

Example: hotel AI agent

A hotel AI agent might use:

AI Answer

For:

• General conversational guidance
• Tone
• Recommendations
• Booking-related instructions

Document Search

For:

• Hotel policies
• Check-in rules
• Cancellation terms
• Guest information

Data Search

For:

• Room types
• Rates
• Availability data
• Property or service records

Each source has a different responsibility.

Example: e-commerce AI agent

An e-commerce AI agent might use:

AI Answer

For:

• Helping the customer choose
• Explaining differences between options
• Conversational recommendations

Document Search

For:

• Returns policy
• Shipping policy
• Warranty information
• Product guides

Data Search

For:

• Product records
• SKUs
• Prices
• Sizes
• Colors
• Inventory

This separation keeps the information easier to maintain.

Example: internal support AI agent

An internal support AI agent might use:

AI Answer

For:

• Internal support behavior
• Guidance on next steps
• Escalation rules

Document Search

For:

• HR policies
• IT procedures
• Internal guides
• Employee documentation

Data Search

For:

• Office locations
• Equipment records
• Structured reference lists
• Other internal datasets

Use the most controlled source for important facts

If important business information already exists in a controlled source, use that source instead of repeating it elsewhere.

For example:

• Product price stored in Data Search → use Data Search
• Cancellation policy stored in Document Search → use Document Search
• Behavior rule → keep it in AI Instructions

Avoid maintaining the same fact in several places unless there is a clear reason.

Duplicated information can become inconsistent over time.

Do not copy large datasets into AI Instructions

Avoid placing long product lists, price tables, inventory records, or lookup values directly into AI Instructions.

These belong in structured sources such as Data Search.

Instructions should explain how the AI agent should use the information, not become the database itself.

Do not copy large document libraries into AI Answer

Similarly, avoid placing entire manuals, policies, or long FAQ libraries into AI Answer when they can be managed as documents.

Document Search is easier to maintain when written knowledge grows.

Ask what kind of question the customer will ask

A useful way to choose is to start with the customer question.

“What does your cancellation policy say?”

→ Document Search

“How much does product X cost?”

→ Data Search

“Which service would you recommend for my situation?”

→ AI Answer

“Is product X available in blue?”

→ Data Search

“What should I do before my appointment?”

→ Document Search or AI Answer, depending on where you maintain that guidance

“Can you explain the difference between these two services?”

→ AI Answer, possibly using information retrieved from another source

Ask how often the information changes

Information that changes frequently should be easy to update.

For example:

Frequently changing structured information

Use Data Search for:

• Prices
• Inventory
• Product catalogs
• Rates
• Reference values

Longer-form knowledge

Use Document Search for:

• Policies
• Guides
• Procedures
• Manuals

Behavioral rules

Use AI Instructions for:

• Tone
• Boundaries
• Required steps
• Decision logic

Keeping these separate makes updates safer.

Ask whether exact values matter

If the AI agent must retrieve an exact field or record, Data Search is usually the better choice.

Examples:

• Exact price
• Exact SKU
• Exact inventory count
• Exact factor
• Exact location ID

If the AI agent needs to understand and explain written content, Document Search is usually better.

If the AI agent needs to guide or reason conversationally, use AI Answer.

A simple decision process

Ask these questions in order.

1. Is this mainly a behavior or instruction?

Yes → AI Answer

2. Is this written knowledge that reads like an article, policy, guide, or manual?

Yes → Document Search

3. Is this structured information organized into records, rows, or columns?

Yes → Data Search

If your use case includes more than one of these, combine the appropriate sources.

Keep each source responsible for one job

A clean setup might look like this:

AI Answer

• Behavior
• Tone
• Conversation
• Skills

Document Search

• Written knowledge

Data Search

• Structured values and records

This makes troubleshooting easier because you know where each answer should come from.

Test each source separately

Before publishing, test questions that should clearly belong to each source.

For example:

AI Answer

“Which service is best for a first-time customer?”

Document Search

“What is the cancellation policy?”

Data Search

“What is the price of product A25?”

Then test mixed questions to confirm the AI agent uses the right information.

Tips for choosing the right source

• Keep behavior and rules in AI Instructions.
• Use Document Search for written knowledge.
• Use Data Search for structured records and exact values.
• Do not choose based only on file type.
• Avoid duplicating the same important information across several sources.
• Use the most controlled source for business-critical facts.
• Combine sources when the use case genuinely needs more than one.
• Test each source with realistic customer questions.
• Keep the information architecture simple enough that your team can maintain it.

Related guides

For more information, see:

• Which AskHandle node should I use?
• How should I write AI Instructions?
• What is a Document Search node?
• What is a Data Search node?
• How can I upload my files?
• How should I prepare documents for better AI answers?
• How should I prepare CSV and Excel files for Data Search?