Learning Center

Migrate from a per-seat help desk without losing history

By the Onsites AI team · Last updated August 28, 2026 · 9-minute read

Teams do not leave help desks for features; they leave over invoices. Per-seat pricing turns growth into a penalty, per-resolution pricing turns every solved ticket into a charge, and "contact sales" pricing makes the exit itself feel dangerous — as if your history lived in their cloud. It does not. History lives in your mailbox and your exports, tags and macros live in your team's heads, and the migration is one quiet day plus a week of parallel running. Here is the plan we walk teams through.

Step 0 · Know what the move is worth

Publish the numbers before you start; it keeps the project honest. Our FAQ compares a typical 10-person support workload at roughly $255/month here versus about $3,100 on a per-seat-plus-per-resolution platform — the difference is fixed credit prices and 3 free seats, not a smaller feature set. The pricing-models post and the comparison page carry the full breakdown. Knowing the number matters for a second reason: a migration you can justify in one sentence does not stall in week three.

Step 1 · Inventory what you actually use

List four things in a shared document: your channels (which of website chat, email, WhatsApp, WeChat, LINE, Telegram or Messenger actually carry volume); your tags (the ten that matter, not the hundred that exist); your macros (the replies your team retypes weekly); and your history (which mailbox, how far back). In our experience half the tags and a third of the macros are dead weight — the migration is exactly the moment to leave them behind. This inventory becomes your cutover checklist.

Step 2 · Connect email first, with history

Email is the spine of every migration because it carries your history. In Settings → Channels → Email choose forwarding (create a shared address, forward the old flow into it — live in minutes) or connect your own domain (verify a DNS record, receive at your real addresses); then use IMAP import to bring the existing mailbox history in, so the last two years of context arrive before your first conversation does. The other channels are minutes each: the widget snippet on your site, your WhatsApp Business number, WeChat and LINE official accounts, Telegram bot, Facebook Page for Messenger — and every one of them lands in the same shared queue.

Step 3 · Rebuild the working parts

An hour or two covers it: paste your top replies in as shared drafts (the whole team gets the good version, not whoever retyped it last); rebuild your working tags (intent tags like refund, shipping-delay double as reports later); set working hours and a first-response SLA so escalation is honest from day one; and invite the team — the first three seats are free, and every seat gets the entire product. Private notes and the auto-linked customer context panel need no setup at all; they are the reason the next step feels smaller than expected.

Step 4 · Run both desks in parallel, then flip

For one week, answer everything in both desks. Forwarding already feeds the new queue while the old tool keeps receiving — nothing breaks, nobody loses a thread, and the team forms the new habit on real traffic instead of a training sandbox. When the counters tell you the new queue is handling it, flip the public-facing addresses: repoint the forwarding rule, update the support email on the site, and let the old tool go quiet as an archive. The cutover is a forwarding flip, not a cliff. Keep the old tool's export files in cold storage regardless — migrations end, archives outlive migrations.

What it costs to try

Nothing. Pilot the entire migration on the free cloud tier — full software, 3 seats, 100 MB, no card. Teams past three seats pay $15/head/month from the fourth; storage beyond 100 MB is $1/GB/month in packs; AI is prepaid credits that never expire. If the pilot convinces you but your regulations require data to sit inside your walls, the same software runs self-hosted per the Docker guide; and if AI is part of the pitch you are moving toward, the honest math is in the monthly AI cost guide. The FAQ answers the rest in plain English.

When the honest answer is "not yet"

Two situations argue for waiting, and both are about contract dates, not capability. A contract that ends within sixty days: run the free pilot in parallel now, learn the queue, and cut over at renewal rather than paying double for a month. A team deep into a contract with heavy customizations: do the inventory and the pilot on the free tier while you wait, then import history and flip on renewal day. Migration rewards calm more than speed; the parallel week exists precisely so the flip lands on an ordinary Tuesday. What never argues for waiting is fear of losing history — mailbox history imports over IMAP, exports preserve the rest, and the worst case either way is an archive folder you would have kept anyway.