TABLE OF CONTENTS
- Prerequisites
- Understand AI agent studio workspace
- AI Agent architecture and operational flow
- Manage AI agent lifecycle
- AI Agent sessions and usage
AI Agent Studio is the centralized workspace to create, configure, deploy, and optimize AI agents across chat and email channels. It enables organizations to move from isolated automation efforts to a governed, measurable AI operating model that improves response speed, consistency, and overall support efficiency.
Prerequisites
Before you begin, ensure that
- You have the required permissions to create and manage AI agents.
- You have defined clear support goals (for example, reduce first response time, increase automation rate, reduce backlog).
- Your knowledge sources - files, articles, FAQs, and workflows are current and structured.
- You have the required licenses for AI capabilities
AI automation amplifies existing processes. Clear documentation and stable workflows are essential before activation.
Understand AI agent studio workspace
AI Agent Studio functions as the control layer for all AI-driven interactions in your support ecosystem. It brings together configuration, behavior control, knowledge grounding, and performance measurement into a single system, enabling AI agents to operate consistently across channels.
The AI agents workspace provides a consolidated view of all configured agents across channels.
- As an Admin, navigate to your left nav bar, and select AI Agent Studio
- You can see a list of pre-configured AI Agents for easy plug-and-play.
- Each agent is represented as a card showing its name, channel, status, and last updated timestamp
- Agents can be filtered, searched, and sorted to manage scale
- The Create option allows you to initiate a new agent configuration
Supported AI Agent types
AI Agent Studio supports two primary execution models.
Chat AI Agent
Chat AI Agents handle real-time conversations across messaging channels such as, WhatsApp, Web Chat, Facebook, Instagram, and more. These agents are designed for quick, real-time conversations where fast response and conversational continuity are critical.
Monitor fallback frequency and repeated replies to detect conversational friction.
Email AI Agents automate responses to inbound support emails. These agents support asynchronous workflows, structured replies, email threading, and escalation logic. They are suitable for high-volume email environments.
Monitor reopen rates and follow-up emails to assess response completeness and clarity.
Chat AI agent vs email AI agent
Capability | Chat AI agent | Email AI agent |
Customer entry point | Web widget or messaging channel — live, real-time conversation | Email inbox — asynchronous, one reply per turn |
Conversation model | Synchronous multi-turn chat session | Asynchronous — each AI reply is a separate email |
Session model | One session consumed per conversation when the first message is sent | No per-session metering — counted per AI reply sent |
Feedback mechanism | Thumbs-up / thumbs-down buttons in the chat widget | Feedback buttons embedded in AI-generated email replies |
Conversation initiation | Agent can greet customers proactively when a session starts | Not applicable — agent responds only after a customer email arrives |
Supported channels | Web widget, WhatsApp, Instagram, Messenger, LINE, and other Freshchat channels | Freshdesk mailboxes, support portals, web forms |
Deployment target | Map to channel — select Freshchat widget or messaging channel | Deploy — select mailboxes, portals, and web forms |
AI Agent architecture and operational flow
AI agents operate through a structured execution sequence that combines business knowledge, behavioral rules, and workflow configurations into a single system.
For every interaction,
- The agent receives a message from a supported channel.
- The agent interprets intent using approved knowledge sources with approved business policies, and reduces the risk of inaccurate or fabricated information.
- The agent executes configured workflows with conditional logic, Actions Library integrations, and confidence-based escalation, working alongside Freshdesk's routing, priority, and queue management systems.
- The agent generates a response based on configured behavioral instructions, including tone, formatting, compliance language, and response boundaries.
- The agent either responds automatically or escalates to a human agent, preserving context.
This architecture ensures that responses are accurate, compliant, and operationally aligned across channels.
Manage AI agent lifecycle
AI Agent Studio is designed around a lifecycle that ensures continuous improvement rather than one-time configuration.
| Stage | What it involves | Why it matters |
|---|---|---|
| Create | Define the agent type (chat or email), supported channels, and success criteria. | Establishes scope and channel alignment |
| Use Library | Leverage reusable components such as prebuilt AI agents, workflows, and API actions. | Accelerates setup, reduces duplication, and helps maintain consistency across multiple agents. |
| Build | Configure the agent using business-approved knowledge, response instructions, and workflow logic such as routing, escalation conditions, and integrations. | Determines how accurately and consistently the agent responds. |
| Test | Validate responses across realistic scenarios, including edge cases, ambiguous queries, and failure paths. Verify tone, accuracy, escalation behavior, and routing outcomes. | Reduces the risk of incorrect or non-compliant responses reaching customers. Address gaps in knowledge, unclear instructions, and workflow misconfigurations before production impact. |
| Deploy | Activate the agent in a controlled rollout, such as limited channels, regions, etc. Monitor early performance signals closely. | Controlled deployment allows teams to detect issues early without affecting the entire support operation. |
| Analyze | Investigate metrics, logs, and patterns to identify issues and opportunities. | Provides visibility into whether the agent is delivering expected outcomes. |
| Iterate | Refine knowledge sources, adjust instructions, recalibrate settings, and update workflow configurations based on observed performance trends. | Drives continuous improvement. Iteration ensures the agent adapts to new customer queries, and changing business requirements, enabling sustained ROI. |
Treat this lifecycle as a recurring loop rather than a one-time process. The highest-performing AI implementations continuously cycle through analysis and iteration.
AI Agent sessions and usage
AI Agent Studio measures usage through sessions, which directly impact capacity planning and cost management.
An AI Agent session includes all interactions between a customer and the AI agent within a 24-hour window.
- A session can last up to 24 hours
- Multiple messages within that period count as a single session
- A session starting late in the day continues into the next day until the 24-hour window completes
This model ensures usage reflects meaningful interactions rather than individual messages.
You can see the session consumption details in the bottom left corner after clicking on AI Agent Studio in the left navigation bar. To add more sessions, click on Buy sessions, which will direct you to the Billing page to purchase additional sessions.
Session FAQs
- What is an AI Agent session?
A session is defined as all interactions between an end-user and the AI agent occurring within a 24-hour window. Each session can last a maximum of 24 hours. - How do I purchase AI Agent sessions?
You can purchase AI Agent Session packs from your billing page or by reaching out to your Account Manager or Customer Success Manager. Each AI Agent session pack contains 100 sessions and costs $49. When you purchase session packs, they will instantly be added to your account for the current billing cycle. For more details, see Manage Freddy AI Ad-ons. - Will previewing AI agent consume sessions?
You can preview your AI agents as you build them, ensuring a well-crafted customer experience. Please be aware that previewing will consume sessions.
Previewing AI Agents will consume sessions, regardless of whether the AI Agent is published or unpublished. - Are there any complimentary sessions available?
Free: No complementary sessions
Growth, Pro, Enterprise: 500 one-time complementary sessions. - What happens when sessions are exhausted?
You will be notified via email when you consume 70%, 80%, 90%, and 100% of your sessions. If you consume 100% of your sessions, the AI agent will stop responding to customer queries, and chats will be handed over to live agents.
If you have deployed your AI Agent using the self-service widget, please note that the AI Agent will stop working, and it won’t be handed over to a live chat agent. - How can I ensure that Freddy AI Agent sessions are automatically added to my account when the available limit is low?
If your account uses online billing, you can enable the Auto-Recharge feature from your Billing page. This automatically adds one session pack to your account when the number of available sessions drops below 50.
Note: The Auto-Recharge feature is not available for accounts with offline billing. - What happens when my customer starts a session at 11:59 PM, which spills into the next day? Will this be counted as two sessions?
Sessions last for 24 hours. If a customer starts a session at 11:59 PM, it is counted as 1 session till 11.58 PM of the following day. - Are unused sessions carried over to the next cycle?
Unused sessions are not carried over to the next subscription cycle.
For example, if you’re on the quarterly billing cycle and use 2000 out of 3000 sessions in a quarter, the remaining 1000 sessions will not be carried over to the next quarter. - How long is the validity of the sessions that I purchased last?
The validity of the sessions purchased aligns with your payment cycle. For example, if you have a quarterly payment cycle, sessions expire quarterly. - What happens when sessions are added in between the billing cycle?
Sessions are automatically added to the current billing cycle and renewed for the next billing cycle. - Does session pricing change according to the chosen billing cycle (Monthly, Quarterly, Half-yearly, Annual)?
No. The pricing remains the same, regardless of the billing cycle you choose. - How can I keep track of how many sessions are used?
Session utilization details can be found in the Session Usage Summary Report. The number of consumed and available Freddy AI Agent sessions will also be available on your billing page. - How can I increase or decrease the number of session packs that I have?
You can increase or decrease session packs based on your usage or support volume from the Billing or Subscription page, or you can reach out to your Account Manager or Customer Success Manager to have them added to your account.
If you need to add additional session packs, they will be added to the current billing cycle. However, if you choose to remove session packs, they will come into effect only in the next billing cycle. For more details, see Manage Freddy AI Ad-ons. - What is the expiry for session packs?
Session packs expire within the billing cycle in which you purchase them. They are not prorated and do not depend on when you add them. They will continue to be added for subsequent billing cycles to ensure that you don't miss out on your requirements during the next cycle.
You must explicitly reduce the session packs during the next billing cycle to avoid getting charged. - Is there any limitation on how I can use my sessions within the billing cycle?
No. If you’re on the quarterly billing cycle and purchased 300 Freddy AI Agent sessions or 3 Packs, these can be used flexibly across the quarter. For example, you can use 100 sessions in the first month and the remaining 200 over the next two months.