How to Add an Email AI Agent to Freshdesk

What an email AI agent does on the ticket channel

An email AI agent works on the ticket that lands in your Freshdesk queue when a customer emails your support address. The moment a new email ticket is created, the agent reads it, classifies the intent, and decides whether it can answer. If it can, it drafts and sends a reply grounded in your knowledge base — no human touch required. If it can't, it hands off to a person.

The important distinction, and one worth holding onto through the rest of this guide, is between three levels of automation. Deflection means preventing a ticket from ever reaching an agent — usually by surfacing a help article. Resolution means the customer's problem is actually solved and the ticket closes. Automation means the agent takes an action on the customer's behalf — checking an order, issuing a refund, updating an address — not just returning text. Most native email AI, Freshdesk's included, is strong on the first, decent on the second for simple questions, and thin on the third. Keep that ladder in mind.

What Freshdesk's native Email AI Agent does well

Freshdesk's built-in option is called the Email AI Agent, part of the Freddy suite. You enable and tune it from the admin console, and for a common class of "where's my article" tickets it works with very little setup.

Per Freshworks' Email AI Agent configuration guide, here is how it fits together:

1. Go to Admin → Freddy → Self Service → Email AI Agent, toggle it on, and click Configure.
2. Set trigger conditions that decide which tickets the agent responds to. The first rule auto-populates as ticket status is Open, and you can layer more with AND/OR logic on group, priority, subject, and custom fields.
3. Manage languages: your account's primary language is the default, and you add others under Manage Languages. A customer writing in a language you haven't enabled simply gets no auto-response.
4. Customise the response template using placeholders such as {{freddy_articles}} (the relevant solution articles it found) and {{freddy_answers}} (a tailored summarised answer), plus standard ticket and contact fields.
5. Set a daily response limit — a safety valve, minimum 10, that caps how many auto-replies go out per day.
6. Choose the applicable sources — Email, Support Portal, Feedback Widget, or Web Forms — then Save & Enable.

Under the hood, per Freshworks' deflection guide, the agent understands the ticket's intent, searches your portal's public solution articles, and returns a summarised reply with up to three relevant articles. It embeds an in-email feedback widget so the customer can say whether it helped, and it auto-closes the ticket when they confirm it did. For a knowledge-base-answerable question, that's a clean deflection with no agent involved. Credit where it's due: it's fast to switch on, it's grounded in your own articles, and it doesn't hallucinate its way into inventing a policy.

How sessions and handoff rules work

Two mechanics govern the native agent day to day: how it's metered, and when it gives up and calls a human.

Sessions. Freshworks documents that every time the Email AI Agent responds, one bot session is consumed. On classic Freshdesk, Email AI Agent sessions are priced around $49 per 100 sessions (roughly $0.49 each), while Omni web-chat sessions run closer to $0.10 each — per eesel's pricing breakdown. An email session is defined as a 72-hour window starting from the customer's first email: every AI reply inside that window counts as one session, no matter how many back-and-forths happen. Plan gating matters too — Growth (Omnichannel) doesn't get the Email AI Agent at all, and on the Growth plan there are no free agent sessions, so you pay from session one. Always confirm the exact numbers against your own plan, since Freshworks revises them.

Handoff. The native agent escalates to a human in three situations, per the Freshworks deflection guide:

1. The customer taps the feedback widget to say the answer wasn't helpful.
2. The incoming email is non-actionable — a newsletter, an out-of-office bounce — so the agent doesn't reply at all.
3. You've built an automation on the "Email AI Agent did not reply" event to route those tickets wherever you want.

That's a sensible, conservative handoff model: when in doubt, a person takes over. It also reveals the ceiling. The native agent hands off precisely when the question stops being answerable from a help article — which is exactly the moment a customer most wants something done.

Where the native agent stops

Be clear-eyed here, because it's the whole reason a layer exists. Freshdesk's Email AI Agent is a knowledge-base responder. It reads the email, matches it to public solution articles, and summarises. That is level-2-to-3 automation: triage and article-grounded assistance. What it does not do is level-4 autonomous resolution with action-taking — the AI fully owning an email ticket end to end, including reaching into other systems to actually complete the request. This is the well-documented boundary of Freddy's email automation, and it's covered in more depth in our Freddy AI explained breakdown.

So when a customer emails "where is my order #4471 and can you change the delivery address," the native agent can find your shipping-policy article. It cannot look up order #4471 in your commerce backend, read the live status, and update the address. It answers about the process; it doesn't do the thing. Everything that requires a real action, a lookup in another system, or multi-step reasoning across your tools ends up back in a human's queue.

Adding the AI agent layer on top

This is the seam where an AI agent layer fits, and it's worth understanding the build-versus-buy tradeoff before you reach for one. The broader category of AI agents for customer service exists to do the reasoning-and-action work a knowledge-base responder can't.

Macha is one such layer. It runs on top of the Freshdesk you already use as a native connector — it does not replace your help desk, and it doesn't turn off Freddy. You connect Macha to Freshdesk with your subdomain and API key, and it reads and writes the same email tickets your team already works. The difference is what happens after the agent understands the email: instead of stopping at an article, a Macha agent can call a custom tool that turns a REST API — your order system, your billing platform, a Postmark send-email endpoint — into an action the agent can take, then reply on the ticket with the real answer.

The screenshot above is the Custom Tools screen from the Macha demo org, not a dedicated email-channel view — but it shows the mechanism clearly: an email AI agent is only as capable as the tools it can call. A send-email tool and an order-shipping-lookup tool are the plumbing that lets an agent resolve "where is my order and change the address" as a single autonomous action rather than a handoff.

Handoff rules on the layer are where you keep control. A well-configured agent should escalate to a human when confidence is low, when the intent touches a sensitive category you've fenced off (billing disputes, cancellations, anything legal), when a required tool call fails, or when the customer explicitly asks for a person. You set those rules deliberately — the point isn't to remove humans, it's to hand them only the tickets that genuinely need judgement. For the fuller picture of wiring automation across the whole desk, see how to automate Freshdesk with AI and our guide to adding an AI agent beyond Freddy.

One honest note on cost model, because it's easy to conflate. Native Freddy meters by session on the email channel. Macha's credits are consumed per AI action — the discrete steps an agent takes — never "per resolution" or "per deflection," because Macha is an automation and orchestration layer and outcomes vary by workflow. That's a different unit of measurement, and you should compare it against your own ticket mix rather than assume parity; the pricing page has the breakdown. (Macha's connector is for Freshdesk specifically — not Freshchat, Freshservice, or Freshcaller.)

The clean division of labour

Put the pieces together and the split is straightforward. Keep Freshdesk's Email AI Agent doing what it's good at: fast, grounded deflection of knowledge-base-answerable email tickets, closed out on positive feedback, at low session cost. Layer a dedicated agent on top for the tickets the native one hands off — the ones that need a lookup, an action, or multi-step reasoning across your tools — with handoff rules that route only the genuinely ambiguous cases to a person. Freddy handles the FAQ; the layer handles the work; humans handle judgement. Nobody's job is to replace the help desk you already run.

FAQ

Where do I set up the Email AI Agent in Freshdesk? Go to Admin → Freddy → Self Service → Email AI Agent, toggle it on, and click Configure. From there you set trigger conditions (the first rule auto-populates as ticket status is Open), enable languages, customise the response template, set a daily response limit, and choose which sources it applies to before you Save & Enable.

How does the native Email AI Agent decide to hand off to a human? It escalates when the customer marks the AI reply as unhelpful via the in-email feedback widget, when the incoming email is non-actionable (a newsletter or out-of-office reply), or when you've built an automation on the "Email AI Agent did not reply" event to route those tickets. When in doubt, it defers to a person.

How is the Email AI Agent priced, and what's a session? Every time the Email AI Agent responds, one bot session is consumed. An email session is a 72-hour window from the customer's first email — all AI replies inside it count as one session. Classic Freshdesk prices Email AI Agent sessions around $49 per 100 (roughly $0.49 each), and the Growth plan gets no free sessions. Confirm current numbers against your own plan.

What can't Freshdesk's native email AI do? It's a knowledge-base responder: it answers from your public solution articles but stops short of autonomous action-taking. It can explain your refund policy but can't look up a specific order, read its live status, or update a delivery address in another system. Those action-heavy tickets get handed to a human.

Can I add an autonomous email AI agent without replacing Freshdesk? Yes. An AI agent layer like Macha connects to Freshdesk as a native connector and runs on top of your existing help desk and Freddy — it doesn't replace either. It reads and writes the same email tickets, and through custom tools it can call your own systems to take an action and reply on the ticket, while you keep Freshdesk as the system of record. Credits are consumed per AI action, not per resolution.

Ready to give your email tickets an agent that acts, not just answers? Start a free trial of Macha and connect it to your Freshdesk in minutes.