> ## Documentation Index
> Fetch the complete documentation index at: https://help.agentsy.pro/llms.txt
> Use this file to discover all available pages before exploring further.

# Prepare Your Knowledge

> Organize your expertise and reference materials into reliable knowledge for your Agentsy AI Experience.

# Prepare Your Knowledge

An effective AI Experience does more than answer generic questions. It helps visitors explore **your** expertise, approach, and services using information you choose to provide.

This guide explains how to prepare that information before uploading it to Agentsy.

> [!NOTE]
>
> **Two different roles:** System instructions guide how your assistant behaves. Your knowledge base supplies the reference information it can use to answer questions. Good results depend on both.

## 1. Start with the questions visitors actually ask

Before collecting files, consider what your audience needs to understand.

* What problems bring visitors to you?
* Which concepts do they regularly misunderstand?
* What questions do they ask before booking a consultation or buying a service?
* What does your approach help them evaluate or decide?
* When should they speak directly with you?

Prioritize content that helps answer these questions. A large collection of unrelated documents is not automatically better than a small, focused knowledge base.

## 2. Identify your knowledge sources

Useful sources may include:

| Source | What it contributes |
| - | - |
| Your methodology or framework | The steps, principles, and reasoning that distinguish your approach |
| Service descriptions | What you offer, who it is for, and what is included |
| Frequently asked questions | Clear answers to recurring visitor questions |
| Educational articles and guides | Explanations, examples, and background knowledge |
| Process documentation | What happens before, during, and after working with you |
| Policies and boundaries | What the assistant should not promise or attempt |

Use only material you own, have permission to use, or are otherwise authorized to provide.

## 3. Separate general knowledge from your unique expertise

A visitor can ask a general-purpose AI to define a concept. Your AI Experience becomes more valuable when it can also explain **how you approach that concept**.

For example, a leadership consultant might prepare:

* **General explanation:** What leadership development involves.
* **Unique methodology:** The consultant's four-stage assessment framework.
* **Practical application:** How each stage helps a client identify next steps.
* **Service boundary:** Which assessments require a human-led session.

Do not present generic AI-generated content as your proprietary methodology. Review and approve the substance yourself.

## 4. Organize content by topic and purpose

A practical starting structure is:

1. **About and approach** — who you help, your philosophy, and how you work.
2. **Core concepts** — essential terms and educational explanations.
3. **Methodology** — your framework, stages, and decision principles.
4. **Services and process** — available services, engagement steps, and expectations.
5. **Common scenarios** — examples that help visitors understand their options.
6. **FAQs and boundaries** — frequent questions, limitations, and when to involve a person.

For complex subjects, create separate documents for individual topics instead of combining everything into one long file.

> [!TIP]
>
> Use descriptive filenames such as `03-leadership-assessment-framework` or `05-common-client-scenarios`. Consistent names make your source materials easier to maintain.

## 5. Write for accurate retrieval

When preparing each document:

* Give it a clear title and a single main subject.
* Use descriptive headings and short, self-contained sections.
* Explain specialized terminology and acronyms.
* State conditions and exceptions close to the claims they qualify.
* Include concrete examples when they improve understanding.
* Remove duplicate, contradictory, or outdated passages.
* Identify the source and revision date for time-sensitive information.

Avoid relying on information that appears only inside complex diagrams, scanned images, or poorly structured tables. If a visual contains important facts, include an accessible text explanation.

### Example: Improve a vague knowledge entry

**Too vague:**

> We offer a personalized process with several stages and ongoing support.

**More useful:**

> Our advisory process begins with an exploratory conversation, followed by a review of the client's goals, a discussion of available approaches, and a recommendation for the next human-led step. The AI Experience can explain these stages but does not conduct a formal assessment or guarantee an outcome.

The second version gives the assistant clearer information and sets appropriate expectations.

## 6. Establish accuracy and safety boundaries

Before uploading, check whether your materials contain:

* Personal, confidential, or client-identifying information.
* Sensitive medical, financial, legal, or employment records.
* Internal pricing, contracts, or proprietary information not intended for visitors.
* Claims that are outdated, unverified, or subject to regulation.
* Instructions that conflict with your intended public-facing experience.

Remove or anonymize sensitive information unless you have a verified, appropriate basis and workflow for processing it. Do not assume that uploading information makes it suitable for public disclosure.

> [!WARNING]
>
> A knowledge base is not a live database. Unless your Agentsy experience has a specifically configured and verified real-time integration, it should not claim to know current inventory, availability, prices, or other changing information.

For regulated topics, clearly distinguish education from professional advice and define when visitors must be directed to a qualified person.

## 7. Prepare the human handoff information

Your assistant should know what happens when a visitor is ready to take the next step.

Document:

* Which requests require a human conversation.
* What your team can help with beyond the AI Experience.
* Which booking or contact options are actually available on your site.
* What the assistant must **not** promise, such as guaranteed appointments or outcomes.

For example, a boat advisor might educate visitors about vessel types and ownership considerations, then suggest speaking with a broker when someone requests current listings. The assistant should not invent boats for sale.

## 8. Upload and test in Agentsy

Once your content is ready, use the knowledge-management features available in your Agentsy account to add the supported documents.

After uploading, test questions that cover:

1. **Direct facts:** Can the assistant explain a service or framework accurately?
2. **Applied questions:** Can it use your methodology in a relevant example?
3. **Missing information:** Does it acknowledge when the materials do not support an answer?
4. **Conflicting information:** Does it avoid confidently repeating inconsistent claims?
5. **Handoff requests:** Does it direct visitors to a person when appropriate?

Review answers against the source documents. If an answer is incomplete or inaccurate, improve the relevant material and test again. Do not assume every answer is grounded simply because documents were uploaded.

## 9. Keep your knowledge current

Assign someone to review the materials whenever your services, policies, methodology, or other important facts change. For time-sensitive content, establish a regular review schedule.

Maintain a simple record of the document owner, last update, and topics affected. Retest representative questions after meaningful changes.

## Knowledge preparation checklist

* [ ] I know the main questions my visitors want answered.
* [ ] My core methodology and services are clearly documented.
* [ ] Documents are organized by topic with descriptive headings.
* [ ] Claims, limitations, and exceptions are accurate.
* [ ] Sensitive or unauthorized information has been removed.
* [ ] Real-time information is not presented as static fact.
* [ ] Human handoff triggers and next steps are defined.
* [ ] I have tested answers against my source material.
* [ ] I have a plan to keep the content updated.

## What's next?

With your knowledge prepared, the next step is to bring your experience's purpose, audience, conversation design, and reference material together into a configuration you can review.

> **[Generate your configuration](/ai-experience-builder/generate-your-configuration)**
>
> Learn how the Builder brings your decisions together.

> **[Knowledge management](/knowledge-management/knowledge-base-overview)**
>
> Explore how knowledge is managed in Agentsy.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.