# Teach Once. Remember Forever.

Train your AI employees on your documents, procedures, and institutional knowledge. They remember everything and never need retraining.

Sistava employees learn from your documentation, SOPs, product specs, and company knowledge. Upload documents, paste URLs, or type instructions. The AI employee absorbs it all, remembers it permanently, and applies it to every conversation. Shared knowledge lets your entire AI team access the same information. Web research fills gaps in real time.

## Overview

Knowledge is what separates a generic AI chatbot from a useful AI employee. A chatbot gives generic answers. An AI employee trained on your documentation gives specific, accurate answers grounded in your actual product, policies, and procedures.

Sistava provides three knowledge layers. Training is the foundation: upload documents, paste URLs, and type instructions to build your AI employee's knowledge base. Memory adds persistence: the AI employee remembers conversations, preferences, and context across sessions. Shared knowledge connects your entire team: information trained into one employee is optionally shared across the team.

For questions that fall outside your training data, AI employees use web research to find current information. They cite their sources, distinguish between your official documentation and web results, and flag when they are uncertain.

## Before / After

- **Before:** Every time you use a new AI chat, you spend the first 10 minutes re-explaining your product, your audience, your tone of voice, and your preferences. It is Groundhog Day with a chatbot.
  **After:** Train your AI employee once with your product docs, brand guidelines, and preferences. It remembers everything across every conversation, forever. No re-explaining needed.
- **Before:** Your best support agent quits and takes 3 years of product knowledge with them. The replacement takes 6 months to reach the same level. During that time, customer satisfaction drops.
  **After:** Your AI employee's knowledge is documented and persistent. It never quits, never forgets, and never needs 6 months to get up to speed. Institutional knowledge lives in the system, not in people's heads.
- **Before:** Your team has documentation in Google Docs, Notion, Confluence, SharePoint, and email threads. Nobody knows which version is current. Your AI tool has no access to any of it.
  **After:** Upload documents from any source. Paste URLs, drag files, or connect cloud storage. Your AI employee indexes everything and answers questions using the most current version of your knowledge.
- **Before:** When your AI tool encounters a question outside its training data, it either makes up an answer (hallucination) or gives a generic "I don't know." Neither response helps the customer.
  **After:** Your AI employee searches the web in real time to fill knowledge gaps. When it cannot find an answer in its training data or online, it says so clearly and offers to escalate. No hallucination.

## Benefits

### Document Training

Upload PDFs, Word documents, spreadsheets, presentations, and text files. Paste URLs to web pages and documentation sites. The AI employee reads, indexes, and retains everything. Updated documents are re-indexed automatically.

### Persistent Memory

Your AI employee remembers every conversation, every preference, and every correction. It learns your communication style, remembers project context, and builds on past interactions. Each conversation makes it more useful.

### Shared Team Knowledge

Training data is optionally shared across your AI team. When you train one employee on your product documentation, the entire team accesses the same knowledge. No duplicate training, no inconsistent information.

### Web Research

AI employees search the web to answer questions that go beyond their training data. They find current pricing, recent news, competitor information, and technical documentation. Results include source citations for verification.

### Knowledge Gaps Detection

The system tracks questions that your AI employee cannot answer from its training data. These knowledge gaps surface in your dashboard so you know exactly what additional training is needed to improve accuracy.

### Source Attribution

Every answer cites the source document, URL, or conversation that informed it. When the AI employee uses web research, it links to the original source. You always know where an answer came from and whether the source is authoritative.

## How It Works

1. **Upload Your Knowledge** — Drag documents into the training panel, paste URLs to documentation sites, or type instructions directly. Supported formats include PDF, DOCX, XLSX, PPTX, TXT, Markdown, and web URLs. Bulk upload is available for large knowledge bases.
2. **AI Employee Learns** — The AI employee indexes your content, understands relationships between documents, and builds a queryable knowledge base. Processing takes seconds for individual documents and minutes for large collections.
3. **Answer with Knowledge** — When a customer or team member asks a question, the AI employee searches its knowledge base first. It synthesizes information from multiple documents, cites sources, and provides specific answers grounded in your actual content.
4. **Continuously Improve** — Review knowledge gap reports to see what questions your AI employee could not answer. Add missing information through additional training. The AI employee immediately incorporates new knowledge without downtime or reprocessing.

## Comparison

| Dimension | Traditional | With Sista |
|---|---|---|
| Knowledge retention | Generic AI starts fresh every session. No memory of past conversations or company context | Persistent memory across all conversations. Cumulative knowledge that grows over time |
| Training effort | Re-explain company context in every prompt. Copy-paste background information into every chat | Train once with your documents. AI employee applies that knowledge to every future conversation automatically |
| Knowledge consistency across team | Each AI instance has different context. Answers vary depending on who set it up and what they pasted in | Shared knowledge base ensures every AI employee gives the same accurate answer to the same question |
| Handling unknown questions | AI either hallucinates a confident wrong answer or gives a generic "I don't know" | AI employee searches the web for current information, cites sources, and clearly flags uncertainty |
| Knowledge updates | Retrain the entire model or update fine-tuning datasets. Changes take hours to days | Upload new documents or update existing ones. Changes are live within seconds |
| Institutional knowledge risk | Knowledge lives in employees' heads. When they leave, the knowledge leaves too | Knowledge is documented, persistent, and organizational. It survives personnel changes |

## FAQ

### What file formats can I use for training?

Upload PDFs, Word documents (DOCX), Excel spreadsheets (XLSX), PowerPoint presentations (PPTX), plain text files (TXT), and Markdown files (MD). You can also paste URLs to web pages, documentation sites, and knowledge bases. The AI employee extracts and indexes content from all of these formats.

### How long does training take?

Individual documents are processed in seconds. Large knowledge bases with hundreds of documents take a few minutes. The AI employee is available immediately; it processes and indexes in the background. You can start asking questions about uploaded content within seconds of uploading.

### Does the AI employee remember previous conversations?

Yes. Persistent memory means your AI employee retains context from past conversations. It remembers your preferences, project context, and previous instructions. This memory grows over time, making the AI employee increasingly useful the more you interact with it.

### Can I share knowledge across multiple AI employees?

Yes. Shared knowledge makes training data available to your entire AI team. When you train one employee on your product documentation, you can enable sharing so all team members access the same information. Each employee can also have private knowledge that is not shared.

### How does the AI employee handle conflicting information in training data?

When documents contain conflicting information, the AI employee prioritizes by recency and source authority. More recently uploaded documents take precedence over older ones. You can also set explicit priority levels for different knowledge sources. The AI employee flags conflicts in its responses when it detects them.

### Does the AI employee make up answers when it does not know something?

No. The AI employee is trained to distinguish between what it knows from your documents, what it finds through web research, and what it does not know at all. When it lacks information, it says so clearly, cites whatever partial knowledge it has, and offers to escalate or research further.