Switch your helpdesk over a weekend: the Friday-to-Monday playbook

Ask a support lead why they're still on a helpdesk they complain about weekly, and the answer is rarely the product. It's the move. A migration feels like a month-long project with a customer-facing blast radius, so it gets postponed quarter after quarter.

Here's the uncomfortable truth: for a typical team of up to ~15 agents, the actual move fits between Friday 6 PM and Monday 9 AM. Not because you rush it — because weekends are when it's cheapest to do. For most B2B products, inbound support volume drops 40–70% on Saturday and Sunday. That lull is a free maintenance window nobody schedules.

This is the playbook we run with teams switching to MoveDesk. It works for any modern destination tool; the principles aren't ours. What you need is one prepared Thursday, one focused weekend, and a rollback plan you hopefully never touch.

Who this works for

The weekend cutover assumes three things:

• Your team is roughly 2–15 agents. Bigger teams don't need more migration time — they need more training time, so they usually spread the same steps across two weekends.
• Your channels are the usual mix: email, website chat, a messenger or two.
• You can freeze helpdesk configuration changes for three days. Not support itself — just changes to macros, automations and articles.

If you're mid-launch, in your seasonal peak, or renegotiating a contract that penalises early exit, don't book the weekend yet. Pick a boring one.

Thursday: the preflight audit

One person, half a day. The goal is to know exactly what exists before anything moves.

The preflight checklist:

• Confirm you can actually export: admin access to conversations, contacts, help-center articles and macros. Some legacy plans hide bulk export behind higher tiers — find out Thursday, not Saturday.
• Inventory your channels: every address that forwards into support, every widget install, every connected messenger. Shadow channels (that old orders@ alias someone set up in 2023) are the #1 source of Monday surprises.
• Count your macros and note the ten your team actually uses. The census matters: teams routinely discover 140 saved replies of which 12 are alive.
• Decide history depth. 12 months satisfies almost every real lookup; 24 is the cautious ceiling. Anything older is archaeology and can import later.
• List integrations, then mark which ones earn money (billing lookups, order status) versus which are habit. Only the first group blocks go-live.
• Name a rollback owner: one person who, at any point during the weekend, can call "revert" and knows the two steps it takes.
• Drop your email DNS TTL to 300 seconds now. If you leave TTL at 24 hours, Monday's switch takes until Tuesday.

Friday evening: freeze and export

From 6 PM, the old helpdesk is read-only for configuration. Agents keep answering tickets; nobody touches macros, automations or articles until Monday.

Then run the exports while nothing is changing underneath them: conversations (CSV or API dump), contacts with custom fields, help-center articles, macros. Kick off the workspace setup on the new side the same evening — team members, roles, brand tone, widget colours. With unlimited seats there's no allocation spreadsheet; everyone gets a login.

Two to three hours, including the pizza.

Saturday: the import day

Order matters more than speed.

Knowledge base first. It's the fuel for the AI agent, and quality beats completeness. This is your once-a-decade chance to prune: an article that answers one real question beats five that half-answer it. Watch for a classic import bug — drafts arriving as published. Spot-check visibility flags before moving on.

Contacts second, with email as the primary key. Map custom fields before the import, not after; a field that arrives as free text instead of a dropdown will haunt your reporting for a year.

History third, in batches, oldest first. Two traps live here: timezone drift (verify a handful of timestamps against the old tool — a silent UTC shift moves every SLA report by hours) and garbage volume. Filter out spam-closed and one-word-closed tickets before export; nobody has ever searched for them.

Macros last. Don't port all 140. Take the ten living ones and convert them into what the new tool does better — AI prompts and suggested replies with the same wording your customers already know.

The old tool keeps answering customers all Saturday. Parallel running isn't a compromise; it's the design.

Sunday: testing and shadow mode

Sunday is proving day, before any customer sees anything.

Replay your golden tickets: pick 25 real conversations from the last quarter — the recurring question, the angry edge case, the refund, the "does this work with X". Send them through the new setup and read every AI answer critically. Does it cite the right article? Does it escalate when it should? Two or three rounds of knowledge-base tuning here is normal, not a failure.

Then the team hour: a 60–90 minute walkthrough of inbox, handoff and article editing. Modern tools don't need training courses, but they do need one guided hour, scheduled — not "you'll pick it up Monday".

Before you close the laptop, everything on this list is green:

• Widget works on staging and matches the brand.
• Email forwarding tested with a real message end to end.
• AI in shadow mode — drafting replies that agents approve, not sending on its own. It stays there for the first production week; the cost of one weekend of review is far below the cost of one confident wrong answer on Monday.
• Rollback rehearsed: old widget snippet restorable in one commit, DNS revert steps written down.

Monday morning: cutover before coffee

7:30 — flip email routing to the new tool. 8:00 — swap the production widget. 8:05 — the rollback owner watches the queue with a coffee.

Tell customers almost nothing. Two sentences: "We've upgraded our support system. Same team, faster answers." Customers care about response quality, not your tooling org chart.

Keep the old tool alive in read-only until the billing cycle ends. Finish in-flight conversations there; start everything new in the new inbox.

What moves, what stays behind

Moves with you: conversation history, contacts and custom fields, help-center articles, tags, the living macros.

Stays behind on purpose: automation spaghetti nobody can explain, custom CSS hacks, saved views from agents who left in 2024. If a workflow only existed to patch the old tool's limitation, recreating it imports the limitation.

Rebuilds later, deliberately: reports (start from the new tool's defaults for two weeks before customising) and nice-to-have integrations.

The four traps

1. Importing everything. Ten years of spam-closed tickets slow the import and pollute search. Filter first.
2. Touching DNS on Monday. TTL goes down on Thursday. Monday you only flip records that already propagate in minutes.
3. Skipping the shadow weekend. Every team that thinks its knowledge base is ready finds three gaps during shadow mode. Better you than a customer.
4. Migrating into a peak. The playbook assumes a quiet weekend. A product launch, a sale, or the Monday your biggest client renews are not quiet weekends.

Where MoveDesk fits

MoveDesk was built around exactly this move — it's in the name. The white-glove migration is free on every plan: our team runs the Saturday import for you — help-center articles, conversation history, macros and settings from Zendesk, Intercom or Freshdesk — while both tools run in parallel for as long as you want to compare. Most teams that start the preflight on Thursday are answering customers in MoveDesk on Monday morning.

The 14-day trial is long enough to run this entire playbook once, on your real data, before you decide anything.

Got a quiet weekend coming up? Start the trial on a Thursday — full features, no credit card required — and we'll handle the Saturday part.