Live Chat Support Team: How to Staff, Schedule and Run It

A live chat support team doesn't start as a team. It starts as a founder answering chats between meetings, then a second person who "helps out," then a moment when chats are slipping through the cracks and nobody is sure whose job it is. This guide is for that moment: how to estimate how many people you need, divide roles, build a schedule, decide who gets which chat, onboard someone new and keep the whole thing running week to week.

Who should answer chat in a small company

In the early days, the best person to answer chat is whoever knows the product best — often a founder. They learn what confuses customers, and customers get expert answers.

As volume grows, the usual options are:

• A dedicated support person — the most common first hire for chat.
• A shared rotation across the team, where everyone takes a shift. Good for keeping everyone close to customers; harder to keep quality consistent.
• Sales and support together, with pre-sale questions going to sales and existing customers going to support. Works well if your chat is mostly leads.

Whatever you choose, someone needs to own chat: watch response times, maintain saved replies and the bot's instructions, and notice when things go wrong.

Estimating how many agents you need: a simple formula

You don't need a workforce management tool to get a reasonable first estimate. You need three numbers:

• Chats per hour at your busiest time — from your chat analytics, or a guess based on traffic if you're just starting.
• Average handling time — how long a typical chat takes from first reply to resolution, in minutes.
• Concurrency — how many chats one agent can handle at once (see below).

Then:

1. Multiply chats per hour by average handling time to get agent-minutes needed per hour.
2. Divide by 60 to get the number of agents fully busy at once.
3. Divide by concurrency.
4. Add a buffer — people take breaks, chats cluster, and some are slow.

Example: 12 chats per peak hour × 10 minutes each = 120 agent-minutes, or 2 agents fully busy. With a concurrency of 2, that's 1 agent. Add a buffer and you probably want 2 people covering the peak.

Treat the result as a starting point and adjust after a few weeks of real data. An AI assistant that answers routine questions first reduces the number of chats reaching humans, so measure handoffs rather than total chats once a bot is in place.

Concurrent chats: how many per agent

Chat's advantage over phone support is that an agent can handle more than one conversation at a time. But there's a limit, and pushing past it shows up as slow replies and mistakes.

Rules of thumb:

• New agents: start with one or two at a time.
• Experienced agents on simple questions: two to three is common.
• Complex technical or emotional conversations: one at a time.

Watch your first response time and CSAT as you increase concurrency. If either worsens, you've gone too far.

Roles and permissions: owner, admin, agent

Not everyone needs access to everything. A clear role structure prevents accidental changes to the widget or the bot's behavior.

In ChatDeck there are three roles:

• Owner — everything, including changing roles and billing. There's one owner per workspace.
• Admin — manages the team (invites and removes agents), the AI assistant, workspace and widget settings, and sees analytics.
• Agent — answers chats and takes conversations from the queue.

A typical small setup: the founder is owner, the support lead is admin, everyone else is an agent. Since ChatDeck has no per-operator fee, you can add occasional helpers — say, a product specialist for technical questions — without thinking about seat costs.

Schedules, presence and coverage gaps

Start with a simple schedule that covers your busiest hours with the most people and your quiet hours with at least one.

Good practices:

• Overlap shifts by 15–30 minutes so open chats get handed over, not abandoned.
• Schedule breaks so two people don't leave at the same time.
• Use presence honestly. Agents set themselves to "away" during breaks so chats don't land with someone who isn't there.
• Publish working hours to visitors, so outside them the widget sets expectations instead of looking abandoned.

ChatDeck's analytics include a team presence timeline showing when each person was online or away, which makes coverage gaps visible. For the hours you can't staff, see our guide to after-hours customer support.

Assignment and queue rules

When chats aren't explicitly assigned, two things happen: two agents answer the same chat, or nobody does. Decide on a model:

• Shared queue, agents take chats. Simple and works for small teams. New chats wait in the queue; an agent takes one and it becomes theirs.
• Specialist routing. Certain topics (billing, technical) go to specific people. An agent can take a conversation or release it back to the queue so the right colleague picks it up.
• Ownership until resolved. Whoever took the chat sees it through, including follow-ups.

In ChatDeck the inbox has tabs for bot chats, the queue, chats assigned to you and resolved ones. Conversations can be open, snoozed or closed, which helps with chats waiting on the customer.

Onboarding a new chat agent in a week

A structured first week gets a new agent productive quickly without throwing them in at the deep end.

1. Day 1 — Product and tools. Use the product as a customer would. Walk through the inbox, statuses and saved replies.
2. Day 2 — Read transcripts. Fifty real conversations, good and bad. Note recurring questions.
3. Day 3 — Shadow. Sit with an experienced agent and discuss replies before they're sent.
4. Day 4 — Reply with review. The new agent drafts answers; a lead checks them.
5. Day 5 — Solo, low concurrency. One chat at a time, with someone available for questions.

Give them a short written guide: tone, response time targets, when to escalate, what never to promise. Our list of live chat best practices is a good starting point.

Running the team: weekly review and feedback

A 30-minute weekly review keeps quality from drifting:

• Look at the numbers. First response time, response rate, wait time after bot handoff and CSAT. ChatDeck shows these alongside per-agent reply counts and response times, compared with the previous period. Our guides to live chat metrics and first response time explain what each one tells you.
• Read a handful of chats — a few low-rated, a few high-rated.
• Update saved replies and the bot's instruction with anything new you learned.
• Give individual feedback privately, with specific examples.

Use per-agent numbers for coaching, not ranking. An agent who takes the hardest chats will always look slower.

Frequently asked questions

How many live chat agents do I need?

Multiply peak chats per hour by average handling time in minutes, divide by 60 and then by the number of chats each agent can handle at once. Add a buffer for breaks and spikes, then adjust after a few weeks of real data.

How many chats can one agent handle at a time?

As a rule of thumb, one or two for new agents and two or three for experienced agents on simple questions. Complex or sensitive conversations are best handled one at a time. Watch response time and CSAT to find your limit.

Should salespeople or support answer live chat?

It depends on what visitors ask. If most chats are pre-sale questions, sales can own them; if most are from existing customers, support should. Many teams split by topic, with a bot routing chats to the right person.

How do I onboard a new live chat agent?

Spend the first week on product knowledge, reading real transcripts, shadowing, reviewed replies and finally solo chats at low concurrency. A short written guide on tone, escalation and promises speeds things up.

Build your team on ChatDeck

ChatDeck is free with unlimited operators, so you can add every person who answers chat without counting seats. Roles, a shared inbox, presence and per-agent analytics come built in. Create an account and see our guide to live chat metrics to set your first targets. To add your teammates and assign roles, follow Invite your team.