Assigning WhatsApp chats to the right agent
A shared inbox where everyone can see everything is a shared inbox where nobody is responsible for anything.
Assignment is what turns a shared inbox from a noticeboard into a workload. Without it, every conversation is technically everybody's and practically nobody's, and the ones that get answered are the ones that happen to be at the top.
There are three ways to decide who gets a chat, and the right one depends more on your team's size than on anything else.
The three modes
| Mode | How it decides | Suits |
|---|---|---|
| Round-robin | Takes turns, in order | Teams whose conversations are broadly similar in size |
| Balanced | Gives the next chat to whoever is carrying least | Teams where some conversations take much longer than others |
| Manual | Nothing is assigned automatically | Very small teams, or ones where a lead triages |
Round-robin is the default, and it is a reasonable one because it is predictable. Balanced is usually the better choice once your conversations vary in length — a complaint and a stock question are not equivalent units of work, and taking turns treats them as if they were.
Availability comes first
Before anything is assigned, the system works out who can actually take it. There are six verdicts, and only the first takes automatic work:
- Available — on shift, present, and under their limit.
- At capacity — working, but already holding as many open chats as you allow.
- Away — they said so, or they went quiet long enough to be marked away automatically.
- Off shift — outside their rostered hours.
- Offline — not signed in.
- Inactive — not a working account.
That distinction matters because "nobody replied" has several different causes, and they need different fixes. An at-capacity team needs more people or a higher limit; an off-shift team needs a rota change.
The capacity limit
You can cap how many open conversations one agent holds at once. It defaults to unlimited, which is fine for a small team and becomes a problem the first time somebody logs in to forty chats.
Setting a limit changes what happens under pressure: instead of one person receiving everything and answering none of it, the surplus is held back. That is almost always the better failure mode, because a queue is visible and an overwhelmed colleague is not.
What happens to chats nobody can take
They queue. When an agent becomes available, the queue drains to them — in batches rather than all at once, so signing in does not hand one person the entire backlog.
Two settings shape this. Whether work is stacked on off-duty agents' queues at all, and how many chats one agent can receive in a single batch. Both are worth setting deliberately, because the defaults are conservative and the right answer depends on how your team actually starts a shift.
Assigning from a flow
An automation can assign a conversation to a person or a team, and that is where most assignment should originate. Two moments in particular:
- When the bot cannot answer — the unanswered path of a knowledge base step, or a menu's fallback.
- When the topic needs a specific person — a wholesale enquiry to whoever quotes, a complaint to whoever can authorise a refund.
Routing by topic beats routing by turn for anything specialised, and the flow already knows the topic because it just asked. The block detail is in internal notes and team alerts from a flow.
Teams, and when they are worth it
Assigning to a team rather than an individual is right when several people can handle the same kind of work and you want whichever of them is free. It is wrong when you are using it to avoid deciding — a chat assigned to a team of eight with nobody on duty is a chat nobody owns.
A workable rule: teams for categories of work, individuals for continuity. A customer already talking to someone should keep talking to them.
What good assignment prevents
| Without it | With it |
|---|---|
| Two people answer the same customer | One owner, visible to everyone |
| Hard conversations get left | They are somebody's, and can be chased |
| No idea who handled what | Workload and response time become measurable |
| New conversations lost under old ones | Each person sees their own queue |
The third row is the one that pays off later. Almost nothing about your team's performance can be measured if conversations have no owner.
Assignment and the clock
One rule worth keeping in mind while you design a rota. The 24-hour customer service window opens when a customer messages you and resets each time, per Meta's sending messages documentation, and inside it replies are free and unlimited under Meta's pricing documentation.
Outside it, reaching the customer costs a template. So a conversation that sits unassigned overnight is not only a slower reply — it is potentially a more expensive one.
Frequently asked
Which mode should a three-person team use?
Round-robin or manual. Balanced starts earning its keep when there are enough people that you cannot see everyone's workload at a glance.
Can a customer be reassigned?
Yes, and it should be, whenever the topic turns out to belong to someone else.
What if someone is on shift but away from their desk?
That is what the away state and the automatic away timer are for. Rostered and present are different things, and assignment respects both.
Does assignment affect the customer's experience?
Only through speed and continuity — they see no labels. The handover mechanics they do notice are in bot to human handover.
Choosing a mode, in practice
Three questions settle it, and none of them are about the software:
- Do your conversations vary a lot in length? If a complaint takes 20 minutes and a stock question takes 2, balanced is the honest choice.
- Can you see everyone's workload at a glance? If yes, manual is fine and gives you the most control.
- Does anyone specialise? Then route by topic from the flow first, and let the mode handle only what is left.
Most teams end up with a combination: flows route the specialised work by topic, and everything general is shared out automatically. That is a better design than picking one mechanism and forcing everything through it.
The settings that decide behaviour under load
| Setting | What it changes | Sensible starting point |
|---|---|---|
| Maximum open chats per agent | When someone stops receiving new work | A number your team recognises as a full plate |
| Assign when away | Whether off-duty agents accumulate a queue | Off, unless your team likes starting with a list |
| Queue batch size | How many chats arrive at once when someone signs in | Small — a batch you can actually work through |
| Automatic away after inactivity | Whether an idle tab counts as presence | On, at a period matching how you work |
| Reassign an away agent's chats | Whether their open conversations move | A team decision, not a default |
Every one of those is invisible on a quiet day and decisive on a busy one, which is why they are worth setting before you need them.
What to do when nobody is available
The queue holds the work, and that is correct — but the customer should not be left with silence. An automation that acknowledges the message, says when someone will reply, and answers anything it can costs nothing and changes the experience entirely.
It is also the difference between a customer who waits and one who messages a competitor. The after-hours version is in what to do with chats overnight.
Rolling it out without upsetting anyone
Switching a team from "everyone sees everything" to assigned work is a change in how people work, and it lands better if you introduce it in stages.
- Week 1 — manual mode. A lead assigns, everyone gets used to seeing a queue that is theirs.
- Week 2 — automate the specialised routing from your flows only. Nothing general changes yet.
- Week 3 — turn on round-robin for everything else, with a generous capacity limit.
- Week 4 — tighten the limit to whatever your team says a full plate actually is.
- Later — consider balanced, if conversation lengths vary enough to justify it.
Four weeks sounds slow and is not: each step is reversible, and the team gets to tell you what the numbers should be rather than being handed them.
The one conversation to have with the team
What happens to a chat somebody cannot finish. That is the question every agent has and almost nobody asks, and the answer needs to be a mechanism rather than a convention — reassign it, or put it back in the queue, but decide which, and make sure everyone knows.
Left undecided, the informal answer becomes "leave it open and hope", and that is where slow conversations come from. 1 clear rule removes a whole category of stalled work.
The short version
Route the specialised work from your flows, share the rest automatically, cap how much one person holds, and let the queue absorb the overflow. Round-robin unless your conversations vary a lot in length, in which case balanced.