Mike HeapMike is an experienced Product Manager who focuses on all the “non-development” areas of My AskAI, from finance and customer success to product design, copywriting, testing and more.
Table of Contents
- What is Freshdesk Freddy AI?
- How easy is it to set up Freddy AI in Freshdesk and Freshchat?
- What channels does Freddy AI work in?
- What are the limitations of Freddy AI?
- What knowledge sources can I train Freddy AI on?
- What features does Freddy AI have?
- How do I improve Freddy AI responses?
- What resolution rate can I expect from Freddy AI?
- What AI model does Freddy AI use?
- What languages does Freddy AI work in?
- How secure is Freddy AI?
- Who is using Freddy AI?
- How much does Freddy AI cost?
- Freddy AI Copilot (per agent)
- Freddy AI Agent (session-based)
- Does Freddy AI have a free trial?
- Is Freddy AI worth it?
- FAQs
Last Updated
Created time
Mar 3, 2026 02:06 PM
Title length (<60)
Author
Ecomm?
Tags
Image
Publish date
Mar 3, 2026
Slug
freshdesk-freddy-ai-agent-complete-guide-2026
Featured
Featured
Type
Article
Ready to Publish
Ready to Publish
I know why you’re here…
Someone (who shall not be named) said “we need to use more AI,” and now you’re holding the bag: you need something that actually works in Freshdesk/Freshchat, doesn’t create a brand-risk dumpster fire, and doesn’t come with pricing that makes your CFO suddenly develop a “customer support is a cost center” philosophy.
There are usually three situations that lead to reading a guide like this:
- You’ve already got Freshdesk and/or Freshchat working nicely, and you want Freddy AI to take real Tier-1 volume off your team (not just write nicer sentences).
- You turned on Freddy AI Copilot, got some value, and now you’re wondering if the Freddy AI Agent is the “real” step-change.
- You’re trying to model cost, and the whole “sessions / packs / add-ons” thing is… not exactly soothing.
This guide breaks down what Freddy AI is, how it works, where it fits, what it costs (as clearly as Freshworks’ own docs allow), what to expect in performance, and how to roll it out without regrets.
What is Freshdesk Freddy AI?
TL;DR: Freddy AI is Freshworks’ generative AI suite for Freshdesk and Freshchat that includes AI Agents (automated chat and email responses), Copilot (agent-assist tools), and Insights (analytics).
Freddy AI is Freshworks’ umbrella name for AI capabilities across Freshdesk and Freshchat. In practical support-team terms, it’s three things:
First, the Freddy AI Agent: this is the “autopilot” piece that can respond to customers directly in chat and email based on your knowledge sources, and can hand off to a human when it can’t (or shouldn’t) answer. Freshworks positioned this more explicitly as an autonomous agent with the launch of the Freddy AI Agent.
Second, Freddy AI Copilot: this is the “agent assist” set of features embedded into the agent experience—drafting replies, summarization, translation, sentiment, and related productivity boosters. Freshworks pushed this hard with the Freddy AI Copilot release and broader generative AI enhancements.
Third, Freddy AI Insights: analytics and reporting intended to help you understand what’s happening (and what’s breaking) across tickets and AI interactions.
Under the hood, the core idea is simple: Freddy learns from what you give it (knowledge base, uploads, URLs, Q&A), uses that to answer common questions, and escalates when needed.
That’s the promise anyway.
How easy is it to set up Freddy AI in Freshdesk and Freshchat?
TL;DR: Freddy AI setup is no-code in Freshdesk’s AI Agent Studio: create an agent, add knowledge sources, configure persona/instructions, optionally add workflows, then map the agent to channels.
Freshworks generally gets the “time to first working thing” right.
Freddy AI Agent setup is designed to be done in the admin UI via Freshdesk’s AI Agent Studio. The documented flow is basically:
- Create the agent
- Choose language
- Add knowledge sources
- Configure behavior/persona
- Assign it to channels
(All described in more detail here: set up AI Agent).
If you keep it simple—knowledge + a standard persona + a clear handoff path—you can get to a working baseline quickly.
Complexity creeps in when you want the agent to do “agentic” things (actions like order lookups, cancellations, subscription changes), because then you’re attaching workflows/skills and you’ll spend time testing edge cases.
Freshworks also supports optional workflow configuration so the AI can take actions, not just answer questions.
That “take action” layer is part of the Freddy AI Agent pitch: connecting the agent to workflows so it can do more than quote a help article.
My AskAI is a direct swap for Freddy AI, with an approved app in the Freshworks marketplace, entirely no-code, you can set it up in under 10 minutes and save 2-5x the cost.
What channels does Freddy AI work in?
TL;DR: Freddy AI supports omnichannel use cases including web chat, email tickets, and messaging channels like WhatsApp, Facebook/Instagram Messenger, and SMS, depending on your Freshdesk/Freshchat setup.
Freddy AI is meant to operate wherever Freshdesk Omni / Freshchat operate.
From the Freshworks docs, you’ll see support for chat-first channels (web widget and messaging) plus email automation:
- Web chat via Freshdesk/Freshchat widgets, with channel mapping documented as part of setting up and deploying agents in Set up AI Agent.
- Email ticket automation via the Freshdesk Email AI Agent, with Freshworks explicitly supporting auto-replies/automation for email in the Freddy suite.
- Messaging channels Freshworks lists for Freshchat include WhatsApp, Facebook Messenger, Instagram, SMS, and others in their messaging channels overview.
One important practical nuance: some channel coverage depends on plan and product (Freshdesk vs Omni vs Freshchat). Freshworks markets “omnichannel” support hard, but in real life you might need to confirm which channels you can enable in your exact plan and environment.
My AskAI works in all of the same channels.
What are the limitations of Freddy AI?
TL;DR: Freddy AI is constrained by knowledge ingestion limits (file size/count and URL limits), only supports public static URLs for web content ingestion, only replying to the first email in a thread, and uses session-pack billing that can create forecasting inefficiency.
Freddy is capable, but it’s not magic, and it’s not super flexible.
The most concrete limitations are around knowledge ingestion and quotas.
They have constraints such as file limits and URL limits in their knowledge configuration documentation, including limits like file size caps and capped numbers of files/URLs per agent/account in Configure knowledge sources.
If your “knowledge base” is actually 400 PDFs and a pile of internal docs in Google Drive/Notion, you’ll begin to see what I mean.
Another practical limitation: URL ingestion is for public, static pages.
If your critical help content lives behind login, or your site is heavy on dynamic rendering, a simple “point at URL” ingestion approach probably won’t work.
And then there’s billing mechanics.
Freddy AI Agent usage is session-based, and Freshworks sells add-ons / packs.
When you have to buy usage in chunks, you can get into the annoying situation where you pay for a block, then don’t fully consume it (or you run out mid-month). ons in Manage Freddy AI add-ons.
The final limitation is that with the email AI agent, Freddy AI only replies to the first message in a thread or chain, not any subsequent messages.
None of these are dealbreakers.
They’re just the boring reality you want to understand before you promise “AI will take 50% of volume next month” to your boss.
My AskAI charges based on ticket volume, so you can more easily predict your usage and keep your costs stable as your resolution increases. You can also connect your My AskAI to more knowledge sources for direct replies, including Google Drive, Notion, OneDrive and more. And My AskAI continues responding on emails, even after the first message.
What knowledge sources can I train Freddy AI on?
TL;DR: Freddy AI can train on Freshdesk knowledge base content, uploaded files, public URLs, and custom Q&A pairs, with documented limits on file size/count and URL count.
Freddy AI’s training sources are pretty straightforward.
You can feed Freddy AI knowledge from:
- Freshdesk knowledge base content (solution articles / FAQs)
- File uploads (like PDFs and docs), with constraints documented in Configure knowledge sources
- Public URLs (again, with constraints: public, static content)
- Custom Q&A pairs (useful for “this question always comes up but we don’t want a full article”)
You’ll also see Freshworks describe the overall setup flow and knowledge source connection in Set up AI Agent.
If you’re thinking “cool, but we need the bot to check an order, update a subscription, or cancel something,” that’s not really “knowledge source training”, that’s the workflows/skills side of the house (covered more in features below).
My AskAI can create tasks, actions and connect to any API to improve responses. You can also connect your My AskAI to more knowledge sources for direct replies, including Google Drive, Notion, OneDrive and more.
What features does Freddy AI have?
TL;DR: Freddy AI includes autonomous AI Agents (chat and email), Copilot agent-assist features (drafting, summarization, translation, sentiment), workflow-based actions, handoff to humans, and analytics/insights.
Freddy AI is not just one feature; it’s a bundle.
At the “AI Agent” layer, the headline capabilities are:
- Automated customer responses grounded in your knowledge sources
- Multi-turn conversation handling (context within a session)
- Branded persona/tone and instruction setting (how the agent “speaks”)
- Human handoff/escalation when needed
At the “Copilot” layer, you get agent-assist tools like reply suggestions, summarization, sentiment signals, and translation, these are covered in their Freddy AI copilot overview.
At the “actions” layer, Freddy can take real actions via workflows/skills—Freshworks explicitly markets the Freddy AI Agent as being able to execute actions via workflows.
It’s the difference between “here’s a help article link” and “I checked your order status and here’s the tracking update.”
And then there’s “Insights,” which is the analytics side.
One feature that tends to matter operationally (because it reduces risk) is how you control out-of-scope behavior and fallbacks.
Freddy supports configuring how the AI behaves when it can’t answer, and you can tune escalation behavior and messaging.
My AskAI has AI tagging capabilities, Insights and analytics to review responses, can reply in internal as well as direct modes and has a self-learning feature to automatically improve AI knowledge.
How do I improve Freddy AI responses?
TL;DR: Improving Freddy AI is mainly a loop of reviewing unanswered/unhelpful queries in the AI Agent analytics, expanding or refining knowledge sources and Q&As, and testing changes before wider rollout.
If you treat Freddy like “set it and forget it,” you’ll get the outcome you probably deserve.
Freshworks’ own improvement guidance is basically: monitor what it can’t answer, monitor what customers flagged as unhelpful, then fix the underlying knowledge or instructions.
This is supported by analytics and agent performance tooling in the AI Agent experience.
In practice, “improve” usually means one (or more) of these:
- Add missing coverage to your KB or custom Q&As
- Clarify existing articles that are too vague (AI can’t quote what you haven’t written clearly)
- Adjust instructions/persona so the AI behaves correctly (especially around edge cases and escalation)
If you’re using workflows/skills, improvement also includes tightening your workflow definitions so the AI reliably collects the right information and executes the right action.
My AskAI allows you to inspect responses to better understand answers, has QA and testing features and also self-learns to fix knowledge gaps.
What resolution rate can I expect from Freddy AI?
TL;DR: Freshworks claims Freddy AI can resolve up to 80% of queries, but published real-world examples range from ~23% to ~75%+ depending on use case, training quality, and what counts as “resolved.”
Everyone wants a single number here. You won’t get one that’s universally true.
Freshworks marketing states Freddy AI can resolve “up to” 80% of customer queries on a well-trained deployment.
But when you look at case studies, you see wide variance:
- Total Expert’s story references deflection around 23% in their deployment, as shown in Total Expert customer story.
- Another example cites resolving over 75% of conversations for UPayments (UPayments customer story).
- Freshworks also publishes customer story outcomes such as improvements in resolution time and channel shift impacts (which are real value, even when “deflection” isn’t the headline metric), like in Woolacombe Bay’s story.
The honest takeaway is: initial rollouts can be materially lower than the “up to 80%” figure, and performance depends heavily on the breadth/quality of knowledge and the complexity of questions.
You’ll want to define what “resolved” means (no handoff? customer didn’t reply? ticket closed?) and measure consistently.
My AskAI has resolved over 1.1m tickets to date with an average resolution rate of 72%.
What AI model does Freddy AI use?
TL;DR: Freddy AI uses a hybrid approach combining Freshworks models with Microsoft Azure OpenAI (GPT) models, with data handling and safety controls described in Freshworks’ AI data/safety documentation.
Freshworks is relatively direct that Freddy uses a mix of in-house capabilities and Azure OpenAI.
The clearest description of model approach and data handling is in the Freddy AI data and safety FAQs, which describes the use of Azure OpenAI for generative tasks and Freshworks’ approach to safety and governance.
If you care about data retention, training usage, and whether your data is used to train models, this is the doc you’ll want to read carefully (and it’s also where opt-out and governance implications are typically addressed).
My AskAI use a suite of different models, namely from OpenAI and Google. We are also constantly running A/B tests and benchmarking to determine whether better quality, faster models can be used.
What languages does Freddy AI work in?
TL;DR: Freddy AI supports 60+ languages for AI Agents, with language coverage varying by specific feature (e.g., chat, email, summarization), and Freshworks publishes per-feature language support details.
If you support customers in multiple regions, language coverage matters a lot—especially if you’re expecting the AI to do real containment rather than “sorry, we only support English.”
Freshworks states Freddy AI Agents can support 60+ languages, and they publish a more granular breakdown of which features support which languages in Which languages are supported by Freddy AI?.
The important nuance is “per-feature” coverage.
It’s common that the core agent reply capability supports a broad set, while some secondary features (certain analytics or specific automation features) may have narrower language support.
My AskAI is multilingual in 95 different languages.
How secure is Freddy AI?
TL;DR: Freshworks positions Freddy AI as enterprise-grade with encryption, role-based access controls, and compliance programs including SOC 2 and ISO 27001, documented on its trust resources and AI trust FAQs.
Security is usually the first question from legal/compliance and the last thing anyone wants to “discover” mid-rollout.
Freshworks maintains a trust portal and public documentation around security and compliance. The AI-specific trust framing (including governance and related considerations) is discussed in Freddy AI Insights and trust FAQs and data handling is covered in the Freddy AI data and safety FAQs.
Freshworks also lists broader compliance and certifications in its trust materials.
My AskAI is SOC-2 Type II certified and GDPR compliant.
Who is using Freddy AI?
TL;DR: Freddy AI is used by Freshworks customers across industries, with customer stories including Total Expert and Woolacombe Bay Holiday Parks describing automation and support efficiency outcomes.
Freshworks publishes a steady stream of Freddy-related customer outcomes.
A couple of referenced examples:
- Total Expert (financial services / CRM context) discusses ticket volume handling, ROI, and deflection outcomes.
- Woolacombe Bay Holiday Parks highlights shifts to digital support and significantly reduced average resolution time using Freshdesk Omni + Freddy Copilot.
Freshworks also references other stories in its customer story library, and you’ll find Freddy in the mix across eCommerce, travel/hospitality, SaaS, and fintech-style support environments.
My AskAI is used by over 200 businesses around the world, with ticket volumes from a few hundred per month up to 100k per month. You can read our recent case studies here.
How much does Freddy AI cost?
TL;DR: Freddy AI pricing includes a per-agent Copilot add-on and a session-based AI Agent add-on sold in packs, with publicly listed rates varying across Freshworks pages and requiring confirmation for your plan/billing terms.
Freddy pricing is where things get a bit… “SaaS-y.”
There are two distinct cost components:
Freddy AI Copilot (per agent)
Freddy AI Copilot pricing is a per-agent add-on, priced at $35/agent/month (monthly billing) and ~$29/agent/month (annual billing).
Because this is seat-based, you can generally control spend by assigning Copilot to only some agents (Freshworks explicitly allows picking who gets Copilot).
Freddy AI Agent (session-based)
This is where automation spend lives, and it’s session-based.
Freshworks describes AI Agent usage being priced in session packs, with new accounts often getting an initial free block (commonly referenced as 500 free sessions) and then paid packs after that in Manage Freddy AI add-ons.
One pricing view shown in Freshdesk pricing references a model like $99 per 800 sessions (roughly $0.12 per session), while another Freshworks doc references $49 per 100 sessions (roughly $0.49 per session).
Both are publicly visible in Freshworks pages:
- Session-pack purchasing is described in Manage Freddy AI add-ons.
- Freddy add-on pricing context also appears in Understanding Freddy AI features and pricing.
What really matters thoug is how sessions are counted.
Freshworks describes a “session” concept for AI Agents (commonly one AI reply to an email, or a time-bounded chat thread), and that definition is what determines whether your bill looks predictable or chaotic month-to-month.
Practical note: Freshworks also sells usage in blocks/packs, which can make forecasts slightly inefficient if you regularly “just barely” tip into the next block.
My AskAI plans start from $199/mo, with additional tickets costing from $0.10 per ticket. This works out around 2-5x cheaper than Freddy. You can compare your ROI for Freddy AI vs My AskAI here.
Does Freddy AI have a free trial?
TL;DR: Freshdesk offers a free trial (commonly referenced as 21 days), and new accounts typically include an initial block of Freddy AI Agent sessions (commonly 500) as part of enabling/testing the add-ons.
Freshworks does support trialing Freshdesk, and Freddy AI can be enabled during trial periods, with AI Agent add-ons being managed as described in Manage Freddy AI add-ons.
Freshworks also documents and references the “first sessions included” idea (commonly 500 sessions for new accounts) as part of the Freddy add-on setup and management experience.
The only “gotcha” is to make sure you understand which Freddy components you’re trialing (Copilot vs AI Agent) and what happens to configuration and billing when the trial ends.
My AskAI has a 30 day, unlimited free trial. Test every feature to your heart’s content and don’t pay a dime. No credit card required.
Is Freddy AI worth it?
TL;DR: Freddy AI is worth it when you have repeatable Tier-1 queries, solid knowledge content to train on, and enough volume for automation savings to outweigh Copilot + session-pack costs.
Freddy AI is best thought of as an automation lever that gets stronger the more repetitive your support demand is.
It tends to be “worth it” when:
- You have a meaningful chunk of repetitive questions (order status, password resets, plan questions, policy questions, setup “how-to”)
- Your knowledge sources are complete and readable (Freddy can only ground answers in what you provide)
- You are willing to run an ongoing improvement loop using the analytics (unanswered/unhelpful → fix KB/Q&A → test again)
Freshworks’ own stories suggest strong ROI and efficiency gains are possible when configured well, as shown in outcomes highlighted in Total Expert’s story and Woolacombe Bay’s story.
The main reasons it’s not worth it are boring but real:
- You don’t have good knowledge content yet (the AI can’t invent your policies accurately)
- You need something that continues to reply to multiple emails in a thread.
- Your tickets are mostly complex, bespoke investigations (AI can help, but containment will be limited)
- Your cost model doesn’t match your channel mix (session-based billing can be fine, but you want to understand it before committing)
One extra, slightly practical point: if you’re evaluating Freddy AI against third-party options, the decision often comes down to whether you want everything native inside Freshworks (Freddy), versus whether you want broader interoperability across helpdesks and knowledge sources via marketplace apps (a common reason teams look elsewhere).
Freshworks is obviously native to Freshworks; that’s both the strength and the lock-in.
If you want a full comparison of My AskAI against Freddy AI, read one here.
FAQs
What’s the difference between Freddy AI Agent and Freddy AI Copilot?
Freddy AI Agent is the autonomous layer that can respond directly to customers in chat and email, trained on your knowledge sources and optionally connected to workflows/actions. Freddy AI Copilot is the agent-assist layer that helps human agents with tasks like drafting replies, summarization, translation, and sentiment, you can read more in the Freddy AI for ticketing overview.
Where do I configure Freddy AI Agent in Freshdesk?
Freddy AI Agent configuration is done through the AI Agent Studio flow: create an agent, add knowledge, configure persona/instructions, then map it to channels, as documented in Set up AI Agent.
What can Freddy AI be trained on?
Freddy AI can be trained on Freshdesk knowledge base content, uploaded files (e.g., PDFs/docs), public URLs, and custom Q&A pairs. Freshworks documents knowledge-source configuration and limits in Configure knowledge sources.
Does Freddy AI support email, or is it only for chat?
Freddy AI supports email automation through Freshdesk’s Email AI Agent functionality (auto-reading and replying/deflecting routine email tickets), and it also supports chat and messaging channels as part of the broader Freddy AI suite described in Freddy AI for ticketing.
Which messaging channels can Freddy AI work on?
Freshworks lists messaging channels including WhatsApp, Facebook Messenger, Instagram, SMS, and others as part of its messaging channel coverage, described in Freshchat messaging channels. Exact availability depends on your Freshdesk/Freshchat/Omni plan setup.
How many languages does Freddy AI support?
Freshworks states Freddy AI supports 60+ languages and publishes feature-level language support in Which languages are supported by Freddy AI?.
What AI model powers Freddy AI?
Freddy AI uses a hybrid approach combining Freshworks models with Microsoft Azure OpenAI (GPT) models.
Does Freddy AI have published security/compliance details?
Yes. Freshworks provides AI trust and security FAQs such as Freddy AI trust FAQs and data governance details in the Freddy AI data and safety FAQs.
What are the main knowledge limitations (files and URLs) for Freddy AI?
Freshworks documents file size and quantity limits as well as URL limits for knowledge ingestion in Configure knowledge sources. Practically, this means you’ll want to prioritize the most important, most-used documentation rather than uploading everything at once.
How is Freddy AI Agent usage billed?
Freddy AI Agent is billed using session-based add-ons sold in packs, and Freshworks documents purchasing/managing these in Manage Freddy AI add-ons. Publicly listed rates appear in multiple Freshworks pages (with different pack price points shown), so you’ll want to confirm the exact rate that applies to your account and billing terms.
What resolution rate should I realistically expect at launch?
Freshworks claims “up to 80%” resolution in ideal scenarios, but published customer examples vary (e.g., ~23% in Total Expert’s story and higher figures cited in other deployments). Real results depend on knowledge coverage, query mix, and how you define “resolved.”
Is there a way to trial Freddy AI?
Freshworks supports enabling and managing Freddy AI add-ons (including initial included sessions) through the add-ons management flow described in Manage Freddy AI add-ons.
