How to build a chat support rota for website chat, WhatsApp and Messenger: coverage, timezones, handoff rules and fair shifts that hold up.
Chat support rarely fails because the team is bad. It fails because nobody knows who is on duty at 7 p.m. on a Sunday, or because a customer on WhatsApp waits while three people each assume someone else has the conversation. A clear shift design fixes both problems, and it matters even more when an AI agent handles the first line and humans take over only when needed. This guide walks through how to size, shape and run a chat rota, and how to keep customers informed when a person is not immediately available.
Start with when customers actually write
Before you draw a single shift, look at when conversations arrive. Many teams design the rota around office hours and then wonder why the busiest moments feel chaotic. Your chat logs already hold the answer.
Pull a few weeks of conversations from your website chat, WhatsApp and Messenger and group them by hour and weekday. You are looking for three things:
- Peaks: the hours when most conversations start, often lunch breaks and evenings after work.
- Quiet stretches: the hours when an AI agent can handle nearly everything and a human only needs to be reachable.
- Handoff timing: not when chats begin, but when they need a person. These are different curves. A chat that starts at 2 a.m. may be resolved by the bot, while a refund dispute that starts at 10 a.m. needs a human right away.
The third point is the one most teams skip. If an AI agent answers routine questions, your human workload is the set of conversations that get escalated, not the total volume. Schedule for that number.
Decide what "covered" means
Coverage is not binary. Define levels before you assign people, so everyone agrees what a good shift looks like.
Three coverage levels
- Live: a person is at their desk, watching the queue, and expected to respond within a few minutes.
- On call: a person is reachable by alert and will respond within a longer, stated window.
- Bot only: the AI agent answers, and anything it cannot resolve is captured for the next live shift.
Most small teams cannot run Live around the clock, and they do not need to. A realistic pattern is Live during your peak hours, On call for the shoulders, and Bot only overnight, with a clear message to customers about what happens next.
Write the levels down and attach a response expectation to each. The expectation is for your team, not a promise you publish unless you can keep it.
Size the team with a simple example
Here is a hypothetical to show the method. Suppose a shop receives 300 chats a month and the AI agent resolves most of them, so only a fraction needs a person. If handoffs cluster between 10:00 and 20:00, you need Live cover for those ten hours, and a single person can often handle that load in one shift. The question then becomes who covers that person's breaks, lunch and days off.
That is the real sizing rule: one seat of coverage requires more than one person. Holidays, sick days, training and breaks mean a single-person rota breaks down within weeks. A common starting point is to plan for at least one backup for each Live slot.
In this made-up example, evening is the busiest block for human handoffs even though many offices would treat it as after hours. A rota built on a standard nine-to-five pattern would miss it entirely.
Build shifts that people can sustain
A rota is a working-conditions document as much as a coverage plan. Shifts that burn people out cost you more than the gaps they were meant to fill.
Practical rules that help
- Keep shift lengths predictable. Four- to eight-hour blocks are easier to staff and to swap than irregular ones.
- Protect rest between shifts. Avoid closing late and opening early the next day.
- Rotate unpopular slots fairly. Evenings, weekends and holidays should be shared, not assigned to whoever is newest.
- Publish early. Give people the next schedule well before it starts, so they can plan their lives.
- Allow swaps. Let colleagues trade shifts, with a simple rule that the swap is confirmed to whoever manages the rota.
- Include non-chat time. Reviewing conversations, improving answers and training the AI agent are part of the job, so schedule them instead of hoping they happen.
A good rota also names a shift lead for each block. This is the person who decides when to pull in a backup if the queue grows, and who is the first stop for tricky cases.
Handle timezones and distributed teams
If you serve customers across the US, UK and Asia-Pacific, a follow-the-sun model can give you wide coverage without anyone working nights. Teams in, for example, Osaka, London and a US time zone can hand over at predictable points.
The danger is the handover itself. Each transition is a place where conversations get dropped. Make it explicit:
- A short written note on open conversations and anything promised to a customer.
- A fixed overlap, even fifteen minutes, where both people are available.
- A rule that a conversation is owned by one named person at any moment, and the owner changes only through a deliberate handover.
Store shift times in each person's own timezone and let the system do the conversion. Spreadsheets that mix local times are a reliable source of missed shifts, especially around daylight saving changes, which happen on different dates in different countries.
- 01Customer asks a question
- 02AI agent cannot resolve it
- 03Rota finds who is on shift
- 04Person takes over and bot pauses
Tell customers what happens next
Even a perfect rota has gaps: a lunch break, an unexpected absence, a busy spike. What customers dislike most is silence, or a promise that is not kept.
When a handoff happens, the message should say what is true. If someone is on shift, say that a person will pick up the conversation. If nobody is, say that the team is offline and when it returns, based on the real rota. Avoid inventing a wait time just to sound reassuring. A guess that turns out wrong damages trust more than an honest "we are offline until 9 a.m."
Good handoff messages share these traits:
- They say a human is involved, so the customer knows the bot is stepping back.
- They state the next step, not a vague "soon".
- They keep the conversation history, so the customer does not repeat themselves.
- They are available in the customer's language when your audience is international.
Review and adjust every month
A rota is a hypothesis. After a month, check it against reality.
- Look at handoffs by hour and compare them with your Live coverage.
- Check unanswered or late conversations and see which hours they cluster in.
- Ask the team which slots felt overloaded or idle.
- Review sentiment and ratings on conversations handled in each block, to see where quality dips.
- Adjust one thing at a time so you can tell what helped.
Also look at what the AI agent could have handled. If humans keep answering the same question in the evening, add that answer to the knowledge base. Every repeated question you fold back into the bot reduces the staffing you need next month.
Doing this in AI SNS CHAT
AI SNS CHAT has scheduling built in, so the rota and the handoff logic live in the same place.
Set up the rota. Go to Helpdesk → Scheduling and set shifts per person and per timezone. Because each person's shift is stored in their own timezone, a distributed team does not need a shared spreadsheet. The Helpdesk section, which includes Email, Scheduling, Translation and Tickets, needs the Standard plan or higher.
Let handoffs route themselves. When the AI agent hands a conversation to a human, the handoff is auto-assigned to whoever is on shift. You do not need someone watching the queue to forward chats by hand.
Send alerts where your team already works. Handoff alerts can go to Slack, Microsoft Teams, LINE WORKS, WhatsApp, email or a webhook. Web push and desktop alerts are also available, and the free desktop app for Windows and macOS adds notifications and an unread badge.
Tell customers the truth. On handoff, the server reads the rota and adds the "what happens next" line itself. It never invents a wait time, which matches the principle above: say what the rota supports, nothing more.
Take over and keep quality high. Conversations from website chat, WhatsApp, Messenger and most other channels appear in Activity → Chat logs. Email lives in the Inbox. In both places, you can take over a chat, which pauses the bot, use saved replies, and use live translation when the customer writes in another language. You can also edit a reply into the ideal answer, which trains the agent so the same question is less likely to need a person next time.
Manage the team. Roles include Owner, Co-owner, Member, Support and Viewer, with per-member permission overrides, so people on shift get what they need without full admin access. Seats depend on your plan: Standard includes 3, Pro 5 and Corporate 10. Workspace settings → Activity records who created, edited or deleted what, which helps when reviewing handovers.
Review. Analytics shows Chats, Topics, Question history and Sentiment. Advanced analytics need the Pro plan. See pricing for current plan details and features for the full list, or connect your channels first.
Conclusion
A good chat rota starts from handoffs rather than total volume, defines what covered means, plans backups, protects the people on it and tells customers the truth when nobody is free. Build it, review it monthly, and keep folding repeated questions back into your AI agent so the human workload shrinks over time.
Frequently asked questions
Do I need a human on shift at all times if I use an AI agent?
Not necessarily. Many teams run Live cover during peak handoff hours and let the AI agent handle the rest, with an honest message about when a person will return. What matters is that customers are told the truth about what happens next.
How many people do I need to cover one chat shift?
At least two for any slot you consider essential, because breaks, holidays and sick days will otherwise leave gaps. Plan a backup for each Live slot and a named shift lead who can call in more help.
How do I handle teams in different timezones?
Store each person's shift in their own timezone and define explicit handover points with a short overlap and written notes on open conversations. In AI SNS CHAT, Helpdesk → Scheduling sets shifts per person and timezone.
Can the system promise customers a response time?
AI SNS CHAT adds a "what happens next" line based on the real rota and does not invent wait times. If you want to publish a response target, base it on what your team has demonstrated it can meet.
Which plan do I need for shift scheduling?
The Helpdesk section, including Scheduling, needs the Standard plan or higher. Check the pricing page for the current plan details, since an admin can change prices over time.


