Routing Forms to Support with AI Auto-Reply
Routing Forms to Support with AI Auto-Reply
By default, a contact form sends an email when someone submits it. You can switch a form so that submissions create real support tickets in the Support dashboard instead — and optionally have AI reply to the customer the moment they submit.
What changes when a form is routed to Support
- Every submission appears as a ticket in Support → Inbox, not as a row under Forms → Submissions.
- The customer gets a magic-link email so they can view the ticket and reply directly.
- The ticket is auto-categorized (e.g. Bug, Feature Request, Account) and an estimated priority is assigned.
- Admins on the project get an in-app notification.
- If AI Reply is turned on for the form, the AI generates a reply right away — either as a draft suggestion for your team or as an auto-sent message to the customer.
- The standard contact-form email (the one that normally delivers submissions to a recipient address) is skipped. Support owns the notification flow.
Set it up from the dashboard
- Go to Forms in the sidebar and create a new contact form (or open an existing one).
- In the form builder, add at minimum:
- A Subject field (text)
- A Description or Message field (textarea)
- An Email field
- Optionally, a Name field
- Open Support → Settings → Forms & AI replies. Your form appears in the Per-form overrides list with a "Not routed" badge.
- Click the form to open its settings. You'll see two sections:
- Field mapping — choose which form field fills the ticket's Subject, Description, Email, and Name. The inputs come pre-filled with the common names, so if your fields use those names you can leave it alone.
- AI Reply — turn on AI replies and pick the mode, tone, length, persona, and any custom instructions.
- Click Save changes. The form is now routed to support — the badge in the list changes to AI draft, Auto-reply, or Uses project default, depending on what you chose.
To stop a form from creating tickets, open its support settings and clear the integration — or just delete the form. Existing tickets the form created are unaffected.
Field mapping
Each ticket has a Subject, Description, Email, and (optional) Name. Field mapping tells the system which of your form's fields fills each one.
- The mapping inputs default to the common field names (subject, description, email, name). If your form's fields already use those names, you're done.
- If your form uses custom field names (something like "feedback_text"), enter that name into the matching mapping input.
- Fields that aren't mapped still travel with the ticket as extra context — they show up in the ticket's metadata so your team can see everything the customer submitted.
- If no field maps to Description, the system fills the ticket body with a "Label: value" list of all the other fields, so the ticket is never empty.
- If no field maps to Subject, the form name is used as the ticket subject.
How the form renders to customers
On your public support page, when a project has a support-routed form, the standard Submit a request sheet renders that form's fields directly — instead of the built-in Subject / Description / Email / Name fallback. Customers fill it in and tap Submit, and a ticket lands in your inbox.
Field types behave the way you'd expect:
- Text — single-line input
- Email — typed email input with validation
- Textarea — multi-line input
- Select / Radio — dropdown or radio group
- Checkbox — toggleable checkbox
- File Upload — drag-and-drop file picker (see below)
File attachments
To let customers attach screenshots, logs, or other files, add a File Upload field to your form in the builder. Customers drag-and-drop files on the public form, and the files arrive on the ticket as native attachments — your team can preview and download them from the ticket sidebar.
A File Upload field can be configured with:
- A max-file-count limit (default 10).
- A per-file size cap (default 10 MB).
- An accepted-type filter, e.g. "images and PDF only", so customers can't drop in something unexpected.
AI Reply on the form
The AI Reply section on each form lets you tune how AI handles tickets coming through that form specifically. Settings include:
- Mode — Draft generates a suggested reply for your team to review before sending; Auto sends the reply to the customer immediately.
- Tone — friendly, professional, formal, casual, empathetic, or technical.
- Length — short (a few sentences), medium, or detailed.
- Persona — free text describing the AI's role (e.g. "Senior support engineer"). This also becomes the message author name the customer sees.
- Language — auto-detect from the customer's message, or pin to a specific language.
- Signoff — a custom closing line, e.g. "— The Acme Team".
- Instructions — hard guardrails the AI follows on every reply, e.g. "Never promise refunds, escalate billing to a human."
- Ask clarifying questions — when the customer's message is too short to act on, the AI asks 1–3 specific follow-ups instead of guessing.
- Ground in help center articles — the AI uses your published help articles as source material, and surfaces the relevant ones as suggested attachments on the reply.
Project-level fallback
If you don't configure AI Reply on a specific form, tickets from that form fall back to your project-wide AI Reply setting (under Support → Settings → AI replies). If neither is configured, the ticket lands with no AI reply.
Tickets that come in via the built-in Submit a request sheet — without going through a custom form — always use the project-level AI Reply setting.
Best practices
- Route a form to Support when its purpose is intake (bug reports, support requests). Keep plain contact forms for newsletter signups, lead capture, or anything that just needs to land in someone's inbox.
- Add a File Upload field when you expect screenshots or logs — it's a much cleaner experience than asking customers to paste links.
- Start AI Reply in Draft mode. Review a handful of drafts, tune the Instructions, then switch to Auto when you're confident in the output.
- Set the Persona to match your team's voice — it's what shows up as the message author, so it becomes part of the customer's brand experience.