AI assistant

An AI assistant is a program built on a large language model that you talk to in a conversation. You write a question or a request, and the assistant replies: drafts an email, summarizes a contract, explains a spreadsheet formula. Examples: ChatGPT, Claude, GigaChat, Alice.

An assistant answers what it's asked. An AI agent gets a task and picks the steps itself: reads data, calls other services, checks the result.

A corporate AI assistant is the same kind of assistant, but connected to company data and rules. It answers from the internal knowledge base, not just from what the model remembered during training, and it runs where the security team allows it. When there's no such tool, employees handle work tasks in personal accounts, and that's how shadow AI appears.

What to compare when choosing a corporate assistant:

• Where data is stored and processed, and who has access to it
• Whether you can deploy the model in your own environment or bring your own encryption keys
• Which company systems it connects to: knowledge base, email, CRM, task tracker
• How an administrator manages access
• What it costs: a per-employee subscription or pay-per-token

In Pachca, you don't need to open an assistant in a separate window. An AI agent is connected as a bot, employees reach it in a thread with a mention, and it takes context from the discussion itself. An administrator decides in the workspace settings who can use the API.