Extending Freshdesk with AI: Options and Limits

Freshdesk can be extended with AI along four routes: Freshworks’ native Freddy AI features, apps from the Freshworks Marketplace, a pre-ticket AI covering the help page, chat and contact form, and custom builds on the Freshworks platform. This article sets out the four, explains how Freddy AI is billed, and describes what the OMQ app does inside the ticket.

How can you extend Freshdesk with AI?

Four routes are available. They are not mutually exclusive, and most teams end up combining two of them.

Route	Where the AI acts	Typical effort	Suited to
Freddy AI	Inside the Freshworks platform	Configuration	A quick start with no additional vendor
Marketplace app	In the ticket or in chat	Installation and connection	Adding capability without replacing anything
Pre-ticket AI	Before the ticket, on the help page, in chat, in the form	Building the knowledge base	Reducing volume before tickets are created
Custom build	Wherever you choose	Development project	In-house systems, specific workflows

Freddy AI from Freshworks

Freshworks groups its AI capabilities under the name Freddy AI. Behind it sit several products: Freddy AI Agent as the customer-facing bot built in AI Agent Studio, Freddy AI Copilot as agent assistance, and Freddy AI Insights for reporting. The appeal is proximity to the platform; the cost is calculated separately from your plan.

Apps from the Freshworks Marketplace

Freshworks explicitly provides for extension. Apps installed from the Marketplace add capability to Freshdesk and Freshdesk Omni. For most service teams this is the least disruptive route, because nothing in the existing configuration changes. The OMQ app sits in this category.

Pre-ticket AI on the help page, chat and contact form

This route deliberately leaves the ticketing system. The AI answers enquiries where they arise, and only what remains open becomes a ticket. The leverage is greatest here, because the enquiry never becomes a case.

Custom builds on the Freshworks platform

Freshworks provides a development kit and APIs for building your own apps and publishing them to the Marketplace, plus AI Actions apps that expose your own operations to AI Agent Studio. This route makes sense when in-house systems have to be involved. It does tie up development capacity on an ongoing basis.

How does Freshworks bill its AI features?

Unlike some competitors, Freshworks publishes its prices. The following comes from the Freshdesk pricing page and Freshworks documentation as of October 2026.

Component	How it is billed
Freddy AI Agent	By session. 500 sessions included once, then 49 US dollars per 100 sessions
Freddy AI Copilot	29 US dollars per agent per month
Freddy AI Insights	Included in the Enterprise feature set

What matters is what Freshworks counts as a session. It is an interaction window, not a resolved enquiry. For the Chat AI Agent, all interactions within 24 hours of the conversation starting count as one session. For the Email AI Agent, a session spans 72 hours from the customer’s first email.

Three further details from the documentation matter for planning:

• Sessions expire with the billing cycle. Pay quarterly and your sessions lapse quarterly.
• Testing consumes sessions too. Previewing an AI agent while building it uses sessions, whether or not the agent is published.
• When sessions run out, the AI agent stops responding and hands conversations over to live agents. In self-service widgets that handover does not happen. Freshworks offers auto-recharge to prevent the interruption.

What does the OMQ app do in Freshdesk?

OMQ Assist is an app in the Freshworks Marketplace, free of charge and available for Freshdesk and Freshdesk Omni. It brings the OMQ knowledge base into the ticket view.

The workflow is short. When an agent opens a ticket, the AI reads the incoming message and determines what it is about. Matching answers from the knowledge base appear in the sidebar. One click inserts the chosen answer into the reply field, ready to send.

Three features matter in daily use:

• Manual search: if the suggestion does not fit, a search field finds other entries. The count shows how many entries exist for that term in the knowledge base.
• Language detection: the ticket language is detected when it opens, and only answers in that language are shown. OMQ works in over 30 languages, including Asian ones.
• One shared source: the answers come from the same knowledge base that feeds the help page, contact form, chatbot and email bot. It is maintained once.

The third point is what separates this from a canned-response library. When a deadline or a price changes, it is updated in one place and takes effect in the ticket and on the public help page at the same time.

How is the OMQ app set up?

Setting it up is an integration and does not touch your existing Freshdesk configuration.

1. Install the app from the Marketplace

Open the Marketplace in your Freshdesk admin area, search for “OMQ Assist” and install it.

2. Generate an API key in the knowledge base

Existing OMQ customers generate an API key for the new connection in the OMQ knowledge base under the Assist module. New users create an account first.

3. Enter the key in the app

The key goes into the app settings. OMQ Assist then appears on the Apps page and is ready to use.

4. Fill the knowledge base

This is where the actual effort sits, not in the technical setup. Starting with the twenty most frequent enquiries works well, and they are already in your Freshdesk reporting.

5. Work in the ticket

Open a ticket, select OMQ Assist in the sidebar, insert the answer. From here the search queries show which content is still missing.

Where AI inside the ticket reaches its limits

The limits below apply regardless of vendor. They follow from the architecture rather than from any product’s quality.

The AI only acts once the ticket exists

An assistant in the ticket speeds up handling but does not prevent the case. Reporting, SLA and handling steps are already in motion. A Gartner survey of 5,728 customers found that 73 per cent use self-service at some point in their service journey, yet only 14 per cent fully resolve their issue there. That gap opens before the ticket.

Without a maintained knowledge base, no model helps

In the same survey, 43 per cent said they gave up because they could not find relevant content, and 45 per cent said the company did not understand their issue. Both causes sit in the content and in intent recognition.

Scattered knowledge produces contradictions

When canned responses in the ticketing system, help centre articles and chatbot content are maintained separately, they go stale in at least one place, and customers get different answers depending on the channel.

The route to a human must stay visible

Gartner found that 87 per cent of customers consider access to a human agent essential once a company uses generative AI in service.

The billing unit is not the success metric

How a vendor bills follows the vendor’s logic. Whether your service improves follows a different one. Sessions, tickets and resolved enquiries are three separate figures.

Extend or replace?

The question comes up in almost every evaluation, and it is usually posed too broadly.

Situation	Sensible response
Handling takes too long but ticketing works	Extend: add assistance inside the ticket
Enquiries arrive faster than they are cleared	Extend: act before the ticket
Automation cost is not predictable	Extend, and model the additional vendor’s billing against your volume
Processes, roles or reporting no longer fit	Evaluate a platform change

Only the last row justifies a migration. The first three are solved by an extension in which no tickets are moved.

Which metrics show the effect?

Metric	What it tells you
Average handling time	Whether assistance in the ticket actually saves time
First contact resolution	Whether enquiries are settled at the first contact
Ticket deflection rate	How many enquiries are resolved before a ticket is created
CSAT on automated conversations	Whether the enquiry was resolved or the customer gave up
Zero-result searches	Which content is missing from the knowledge base

Deflection and CSAT belong in the same report. Falling ticket counts alongside falling satisfaction is not a saving, it is displacement into channels nobody measures.

What should decide the selection?

Seven questions any extension vendor should be able to answer:

1. Does the solution act before or after the ticket? That determines how much volume it can reach.
2. What source does it answer from? Ideally one knowledge base serving every channel.
3. What exactly triggers a charge? Session, conversation, resolved enquiry or seat. The models behave very differently as you grow.
4. Does it understand free-text enquiries? Paraphrasing, typos and synonyms, not just keyword matching.
5. Which languages are supported, and is the ticket language detected automatically?
6. Can the knowledge base be exported? Ask about format and scope.
7. What compliance evidence exists? Processing agreement, processing location, subprocessors, retention periods.

What do GDPR and the EU AI Act require?

A ticketing system processes personal data, and so does every AI connected to it. UK GDPR and EU GDPR therefore require a processing agreement under Article 28 with each vendor involved, named subprocessors and transparency about where data is processed. The whole chain needs checking, not just the help desk vendor.

Since 2 August 2026, Article 50 of the EU AI Act has also applied: people must be able to tell that they are interacting with an AI system, unless that is already obvious from the circumstances. For chatbots the duty applies with no transition period. An assistant that only suggests replies to agents falls outside it, because a person sends the message.

OMQ is a German company based in Berlin and provides a processing agreement under Article 28 GDPR including a named external data protection officer. Personal data is stripped out of free-text fields before the text is stored. Further details for tender responses are available through the contact form.

Key terms at a glance

• Freddy AI: Freshworks’ umbrella name for its AI capabilities, comprising AI Agent, Copilot and Insights.
• Session: the billing unit for Freddy AI Agent. An interaction window of 24 hours in chat or 72 hours by email, regardless of outcome.
• Freshworks Marketplace: the directory used to install apps into Freshdesk and Freshdesk Omni.
• FDK: the Freshworks development kit used to build apps for the Marketplace.
• Ticket system assistant: an extension that suggests answers from a knowledge base inside the open ticket.
• Ticket deflection rate: share of enquiries resolved before a ticket is created.
• Intent recognition: the method an AI uses to determine the purpose behind a free-text message rather than matching keywords.

Conclusion

Extending Freshdesk with AI is usually the smaller and more effective move than changing platform. Four routes are open and they combine.

When weighing them up, look closely at the billing unit. Freshworks bills the AI Agent by interaction window rather than by resolved enquiry, and sessions expire with the billing cycle. That is documented transparently, but it needs active monitoring so the allowance does not lapse mid-month.

The OMQ app works at a different point: it speeds up handling inside the ticket and draws on the same knowledge base that feeds the channels in front of it. The biggest lever still sits earlier. An enquiry resolved on the help page costs neither a session nor handling time.

Frequently asked questions

How can you extend Freshdesk with AI?

What is the OMQ app for Freshdesk?

How does Freshworks bill its AI features?

What counts as a Freddy AI Agent session?

What happens when Freddy sessions run out?

How is the OMQ app installed in Freshdesk?

Which languages does the OMQ app support?

Where does AI inside a ticketing system reach its limits?

About OMQ

OMQ is AI software for automated customer service. It is operated by OMQ GmbH, a SaaS company founded in Berlin in 2010 by Matthias Meisdrock and Sven Engelmann. All products draw on the same central knowledge base: AI Chatbot, AI Email Bot, AI Voice Bot, AI Agent, Intelligent Help Page, Smart Contact Form and Ticket System Assistant. Customers include Deutsche Bahn, Tchibo, Mister Spex and alltours.

Current as of 7 October 2026. Statements about Freshworks reflect its public pricing page and documentation at that date and are subject to change. This article is reviewed at least quarterly.

Sources

• Freshworks: Freshdesk Pricing & Plans
• Freshdesk Support: Purchase and Manage Freddy AI Add-ons in Freshdesk
• Freshdesk Support: Introduction to AI Agent Studio
• Freshworks Marketplace: OMQ Assist
• Freshworks: Developer Documentation
• Gartner: Only 14% of Customer Service Issues Are Fully Resolved in Self-Service, survey of 5,728 customers
• Gartner: 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent