How to Reply to Freshdesk Tickets Using Knowledge Base Articles (Automatically)

The fix is to make the knowledge base do the replying. When a ticket comes in that matches an existing article, the article's content should become the reply — drafted, formatted, and either suggested to the agent or sent automatically. This guide walks through every way to do that in Freshdesk: the built-in Suggested Solutions feature, Freddy AI, and an AI agent layer (Macha) that drafts the full reply, links the article, and updates the ticket — pulling from your Freshdesk articles and knowledge that lives outside Freshdesk.

The three native Freshdesk paths (and where each stops)

Freshdesk ships with three escalating levels of knowledge-base-to-reply automation. Knowing which one you're actually on matters, because they price and behave very differently.

1. Suggested Solutions — manual, free, keyword-based

The oldest and simplest. Freshdesk reads the subject and description of a new ticket, pulls keywords, and searches your published solution articles for matches. The agent sees a list of suggested articles next to the ticket and can insert the link or the full article text into the reply with one click.

This is keyword matching, not AI — it doesn't understand intent, it doesn't rephrase, and it only searches published articles. But it's included on paid plans and it's a genuine time-saver for the long tail of "the answer is literally an article" tickets. The agent still writes and sends the reply; Suggested Solutions just hands them the raw material.

2. Freddy AI Reply Suggester — drafts the reply, first message only

Freddy AI Copilot goes a step further. Its Reply Suggester (the "Reply using AI" button) reads the ticket, finds relevant knowledge base content — including public, private, and bot articles — and drafts a full reply in your brand voice, citing the source articles so the agent can validate before sending.

Two things to know. First, it requires Solution Article Suggester to be enabled to show sources. Second — and this is the catch most teams hit — the Reply Suggester only fires on the first customer reply in a ticket. Follow-up messages in the same thread don't get an AI draft. Freddy AI Copilot runs roughly $29 per agent per month as an add-on on top of your Freshdesk plan (Pro is ~$49/agent/mo, Enterprise ~$79/agent/mo on annual billing; verify current numbers on the Freshworks pricing page).

3. Email AI Agent — fully automated deflection

The top tier. Freshdesk's Email AI Agent uses generative AI to answer email tickets directly from your knowledge base without a human in the loop, and can take post-response actions like tagging or escalation. This is the one that actually closes tickets.

It's priced by session, not by seat. Pro and Enterprise plans include 500 free sessions (once per account) to try it; after that the Email AI Agent on Classic Freshdesk runs $49 per 100 sessions — about $0.49 per session — where a session is a 72-hour window from the customer's first email, covering all AI replies inside it. The cheaper figure you'll also see quoted — $100 per 1,000 sessions (~$0.10/session) — is the web/chat agent (Freshbot on Freshdesk Omni), a different product, so don't budget the email use case at chat rates. Both numbers come from eesel's Freshdesk Freddy AI pricing breakdown; session math gets surprising at volume, so confirm before you commit.

The shared limitation. All three native paths read from the Freshdesk knowledge base only. If half your real answers live in Confluence, Google Docs, a Notion runbook, or an internal wiki, Freddy can't see them. That's the gap an AI agent layer is built to close.

The agent approach: draft from any knowledge source, then act

Macha is not a Freshdesk replacement — it's an AI agent layer that runs on top of the Freshdesk you already use. You connect Freshdesk with an API key, point Macha at your knowledge (Freshdesk articles and Confluence, Notion, Google Docs, uploaded files, or a crawled help center), and build an agent that triggers on new tickets, finds the right answer, and drafts the reply — with the option to send automatically or leave it as a draft for an agent to approve.

The difference from native Freddy isn't the AI quality — it's the scope of knowledge and the actions on the back end. A Macha agent doesn't just draft text; it can post the public reply, set the status, apply tags, and add an internal note, all in the same run. And it works on every reply in a thread, not just the first.

Step 1 — Connect Freshdesk

Add Freshdesk from the connectors screen using an API key. Once connected, the agent gets the full Freshdesk toolset: Get Ticket, Search Tickets, Search Articles, Add Public Reply, Add Internal Note, Update Ticket, and Search Contacts.

Step 2 — Point the agent at your knowledge

This is where the agent approach pulls ahead. In Macha, a knowledge source can be your Freshdesk solution articles, a Confluence space, Notion pages, Google Docs, an uploaded PDF, or a crawled public help center — indexed together into one searchable layer. The agent searches all of them when it drafts a reply, so an answer that only exists in an internal Notion doc still makes it into the customer-facing response.

Step 3 — Write the agent instruction

The agent's behavior is plain English. Something like:

When a new Freshdesk ticket is created, read the customer's question, search the knowledge base for the most relevant article, and draft a clear public reply with step-by-step instructions and a link to the full article. If the match is confident, set the status to Pending and tag kb-article-sent. If nothing relevant is found, leave the ticket untouched for a human and add an internal note saying no KB match was found.

That last sentence is the important one — it's the honesty guardrail. The agent only replies when it actually has a confident article match; everything else stays human.

Step 4 — Watch a run

Here's the worked flow from a real reply agent. A customer submits:

"How do I set up SSO with my identity provider? We use Okta and need to configure SAML."

The agent searches the knowledge base, finds "Setting Up SAML SSO with Okta" at a 96% match, and drafts:

Great question! Here's how to set up SAML SSO with Okta: 1. In your admin panel, go to Settings → Security → SSO Configuration 2. Select "SAML 2.0" as the authentication method 3. Copy the Entity ID and ACS URL into your Okta app settings 4. Upload the metadata XML from Okta back into the SSO settings 5. Enable SSO and test with a sample login Full guide: [Setting Up SAML SSO with Okta] Let me know if you run into any issues during setup!

Then it sets the ticket to Pending and tags it kb-article-sent, sso-setup. One ticket, drafted and routed, no agent keystrokes — and because the reply cites and links the article, the customer can go deeper if they need to.

Native Freshdesk vs. an agent layer

The honest read: if your answers all live in Freshdesk and you mostly need first-reply suggestions, native Freddy is the simpler buy — it's one toggle and an add-on line on your existing invoice. The agent layer earns its place when your knowledge is spread across tools, when you need automation on follow-ups, or when "reply" needs to come bundled with real ticket actions and routing.

Pricing: seats and sessions vs. actions

Freshdesk's AI is priced two ways at once — per agent for Copilot (~$29/agent/mo) and per session for the Email AI Agent ($49 per 100 sessions, ~$0.49/session, after the free 500). The web/chat Freshbot is a separate, cheaper line at ~$100 per 1,000 sessions ($0.10). That's predictable if your volume is flat and surprising if it spikes.

Macha prices by credit per AI action, not per seat or per resolved ticket. A reply-drafting run consumes credits for the model actions it takes (0.5–9 credits depending on model, with the default GPT-5.4 Mini at 1 credit), so a simple KB lookup-and-draft is cheap and a multi-tool run costs proportionally more. Crucially, credits are per action, not per deflection — Macha is automation and orchestration, and outcomes vary, so we don't pretend to bill by a "resolution" we can't guarantee. Start with the 7-day free trial, no credit card required and see the live plans on the pricing page.

When NOT to automate this

A few honest watch-outs before you turn anything on:

• Thin or stale knowledge base. If your articles are out of date or sparse, automated replies will confidently send wrong answers. Fix the KB first; an agent is only as good as what it reads.
• High-stakes or regulated topics. Billing disputes, security incidents, legal, and account-deletion requests should stay human or run draft-only with mandatory review.
• Ambiguous questions. When a confident match isn't found, the correct behavior is silence plus a human handoff, not a guessed reply. Build that into the instruction (as in Step 3) and the agent stays trustworthy.
• Tone-sensitive moments. An angry customer doesn't want a polished KB excerpt — they want acknowledgment. Route negative-sentiment tickets to a person.

Start in draft mode — agent writes, human approves — for a week or two, watch the drafts, then graduate the safe categories to auto-send. That's the path that keeps CSAT intact.

FAQ

Can Freshdesk automatically reply to tickets from knowledge base articles? Yes — three ways. Suggested Solutions hands agents matching articles to insert; Freddy AI's Reply Suggester drafts a reply from your KB (first customer message only); and the Email AI Agent fully auto-resolves email tickets from KB content without a human. Coverage and price increase at each level.

Does Freddy AI read knowledge outside Freshdesk? No. Freshdesk's native AI reads the Freshdesk knowledge base only — public, private, and bot articles. Answers living in Confluence, Notion, or Google Docs are invisible to it. A layer like Macha indexes those sources alongside your Freshdesk articles so the agent can draft from all of them.

How much does Freshdesk's AI reply feature cost? Freddy AI Copilot is roughly $29/agent/month as an add-on; the Email AI Agent is session-based — 500 free sessions on Pro/Enterprise, then $49 per 100 sessions ($0.49/session) on Classic Freshdesk. The web/chat Freshbot is a different product at about $100 per 1,000 sessions ($0.10), so don't mix the two when budgeting. Confirm current pricing on the Freshworks pricing page and eesel's pricing breakdown, since session definitions and rates change.

How is Macha different from Freddy AI? Macha is an AI agent layer on top of Freshdesk, not a replacement. It drafts replies from Freshdesk and external knowledge sources, works on every reply in a thread, and can take ticket actions — set status, tag, add internal notes — in the same run. Pricing is per AI action (credits), not per seat or per session.

Will the AI send wrong answers? Only if you let it. Build the agent to reply only on a confident article match and hand everything else to a human, run in draft-only mode first, and keep your knowledge base current. With those guardrails, automated replies stay accurate.

Try it

If your team is still copy-pasting solution articles into Freshdesk by hand, you're paying agents to do what an agent could. See the Freshdesk integration, browse more Freshdesk use cases, or start a 7-day free trial, no credit card required and build a reply agent against your own knowledge base in an afternoon.

Written by Abbas (Customer Support & AI, Macha) · Reviewed by Ankeet Guha (Co-founder & CTO) · Published 2026-06-24 · Last updated 2026-06-24.