Introduction to Zenphi AI Studio

Welcome to Zenphi AI Studio, your dedicated environment for designing, governing, and deploying custom business AI Agents.

While generic large language models (like ChatGPT or Gemini) are excellent at general knowledge, they know nothing about your company. Zenphi AI Studio solves this by allowing you to build intelligent, conversational agents that are securely tethered to your proprietary enterprise data, automated workflows, and internal business processes in a governed and audited way.

In simple terms, AI Studio lets you create AI agents that can start Zenphi flows through natural language and answer questions using the data sources you configure. Your staff can use Google Chat, Zenphi Assistant, or other configured channels to ask questions like, “Do I have any pending tasks?” or “What is the status of my last expense claim?” They can also make requests such as, “I need to send an NDA to John Doe.” The agent can then ask follow-up questions and gather the information needed to submit the NDA request, assuming you have configured an NDA agent.

These are not just chatbots that answer questions—they are active participants in your business that can get the job done by executing workflows through natural language.

Who Can Use AI Studio

While in Preview, AI Studio is available to all Zenphi subscribers. Before AI Studio leaves Preview, we will publish availability details for each Zenphi subscription plan.

Access and interaction within Zenphi AI Studio are defined by clear user roles:

• Agent Management Panel Access: The AI Studio designer is available to Workspace Admins and Designers.
• End-User Availability: Published agents can be made available through channels such as Zenphi Assistant and Google Chat, including to Viewers, depending on the agent’s visibility and security settings.
• Interactive Capability: End users interact with published agents through those deployed channels, subject to the visibility filters, row-level filters, and column exposure configured by the agent designer.

The Core Concept: Purpose-Built Micro-Agents

The guiding philosophy behind Zenphi AI Studio is the Micro-Agent Architecture. Rather than building one broad, all-purpose AI agent to handle every department’s needs, AI Studio is designed around specialized, purpose-built agents that execute tasks in a controlled way.

You might create an “IT Helpdesk Agent” to reset passwords, a “Time-Off Agent” to check leave balances and request for new ones, and an “Expense Agent” to process receipts. This focused approach improves accuracy, strengthens security boundaries, and creates a cleaner user experience.

The Anatomy of an AI Agent

Before you dive into the actual creation and configuration steps, it helps to understand the core building blocks of a Zenphi AI Agent. When you build an agent in the Studio, you will navigate through four primary configuration pillars:

1. Data Integration (The Agent’s Brain)

This is where you connect the agent to Zenphi Tables and internal knowledge bases. This grounds the AI in reality, allowing it to provide factual, company-specific answers rather than guessing or hallucinating.

2. Action Routing (The Agent’s Hands)

AI should do more than just talk. In this section, you connect the agent to specific Zenphi Flows. This empowers the agent to take real-world actions on behalf of the user, such as generating an invoice, updating a CRM record, or routing an approval request.

3. Security & Governance (The Agent’s Boundaries)

This is where you define the agent’s access boundaries. You configure Agent Visibility, Row-Level Filters, and Column Exposure so users can only access data they are allowed to see.

4. Prompt Configuration (The Agent’s Personality & Rules)

Finally, you establish the agent’s operational rulebook. By defining the Agent Description and Behavioral Guidelines, you tell the AI exactly what its job is, what tone it should use, and how it should handle edge cases or missing information.

Example in Action: The “Expense Agent”

To visualize how these pillars come together, consider a standard use case:

• The Mission: You build an “Expense Agent” designed solely to help employees check and submit expense reports.
• The Brain & Hands: You connect the agent to your Expenses table and an Approve Expense flow.
• The Boundaries: You apply a Row-Level Filter ensuring an employee can only see their own expenses, while a manager’s attribute allows them to see their entire team’s pending requests.
• The Experience: Instead of navigating complex software portals, an employee simply opens the chat and types, “What are my pending expenses?” or “Submit this $50 receipt for client lunch.” The agent securely fetches the correct, filtered data or triggers the right flow instantly.

Deployment Channels: Meeting Users Where They Are

Building a useful AI agent is only part of the experience; making it available where your team already works is just as important. Zenphi AI Studio does not force your users to log into a new, unfamiliar portal. Instead, you can deploy your finalized agents directly into the communication platforms your team already uses every day.

Currently, you can publish your Zenphi AI Agents to the following channels:

• Zenphi Assistant: A secure, native chat interface provided by Zenphi, available at zenphi.ai.
• Google Chat: Seamless integration for Google Workspace organizations, allowing users to ping the agent like a regular colleague.

Additional channels are on the way:

• Microsoft Teams (coming soon): Direct deployment into Teams channels or direct messages for organizations running on the Microsoft ecosystem.
• Slack (coming soon): Direct integration into your Slack workspace channels and direct messages.
• Future Platforms: The Zenphi ecosystem is continuously expanding, with plans to support even more communication channels and ticketing systems in upcoming releases.