Auto-assignment and routing: stop fighting over tickets

By the Onsites AI team · Last updated August 28, 2026 · 4-minute read

Someone has to look at every conversation, and on a team larger than one, "someone" is the most expensive word in support. Left alone, queues develop free-rider dynamics: the conscientious carry the slack, the busy miss the urgent, and everyone loses an hour a week to "was this yours?". Auto-assignment fixes the distribution; routing rules fix the fit. Neither takes more than a few minutes to set up on a free workspace — and both should start small.

The three assignment models, honestly compared

First-touch claiming is the default state of a shared queue: whoever reads a conversation first claims it. Its virtues are real — zero configuration, instant self-balancing on a chaotic day, and it teaches everyone the whole surface of the desk, which matters enormously in a company of three. Its vice is the free-for-all: the first reader of the morning inherits the worst thread of the week, and unclaimed items age silently unless a ritual — the end-of-day sweep from the shared inbox guide — sweeps for them. First-touch is the right start for teams of two or three who sit within earshot of each other.

Round-robin deals each new conversation to the next person in sequence. Loading is even by construction, idle time shrinks, and nobody re-learns the queue layout every morning — the desk routes itself. Its costs: the assignment ignores expertise (the invoice question can land on the person least able to answer it), uneven difficulty between tickets makes "even" an illusion, and if someone is out sick the ruleset must know. Round-robin fits teams of three-plus with steady volume and similar work. Most desks that switch to it then spend their first week tuning one exception: a rule that hands specialized topics to their owner.

Topic owners route by tag — billing to whoever owns billing, shipping to whoever owns the carrier relationship. This is the best-quality assignment and the most fragile: it depends on tags being honest, which is why the tagging guide is so fussy about vocabulary. Use topic routing for the two or three areas where expertise really concentrates, and let everything else fall back to round-robin.

Channel rules: one inbox does not fit all channels

Different channels carry different temperatures. A website chat conversation is usually pre-sales or a quick question — fast, short-lived, often better handled by whoever is free right now. Email carries complexity — complaints, refunds, disputes that deserve the person who knows the account. Messaging channels like WhatsApp skew personal: one customer, one relationship, often one preferred agent. Set the channel-specific behaviors accordingly: chat gets the "claim within minutes" norm, email gets round-robin or owner routing, and the high-stakes channels get a first-touch norm with a visible unclaimed warning. The omnichannel guide explains why one shared queue behind all of them — rather than per-channel silos — is what makes these distinctions manageable at all.

The fallback rules that prevent unclaimed queues

Every assignment scheme needs a "no one home" path, or you will meet it as a Monday morning surprise. Write these three before writing anything clever:

• Absence handling. When someone is on leave, auto-assignment must skip them — not because the software will misbehave, but because a manually-managed skip list is exactly the kind of thing everyone forgets the morning after the flight departs. Tie absences to the working-hours settings so both the SLA and the routing pause together.
• Reassignment on silence. If an owned conversation ages past a threshold with no reply, it should return to the unclaimed view and announce itself at the daily check-in. This is the safety net under every model above.
• The overflow rule. When today's volume is double the team, the honest overflow is a stated expectation, not a hidden backlog: "we're at twice normal volume, replies will take until tomorrow morning — here's what we'll do first." Sent by a person, it halves the angry re-messages; the SLA guide shows how to detect such days before lunch.

What routing must never do

A short list of anti-patterns from small teams that automated themselves into corners. Don't route by keyword alone — "invoice" in a subject line is not intent; a customer typed "invoice" to ask about a shipment. Keyword rules over-fit to last month's wording; tags set by humans at close age far better. Don't create a queue per product — it re-creates the silo problem shared inboxes exist to solve, and cross-product questions, which are most of them, get trapped between queues. Don't hide the manual override: the point of rules is to save clicks, not to remove judgment — any assignment scheme your teammates cannot override in one click will be worked around, and circumvented routing is worse than no routing. And don't assign work to the absent — see rule one, which teams violate precisely once.

When to change the model

The queue tells you when switching, the same way traffic tells you when to add a lane. "Who has this?" asked twice a day means first-touch has become a bottleneck — move to round-robin. "Nobody can answer this but Sam" said twice a week means expertise is concentrated — formalize topic ownership with tags. And if conversations routinely get reassigned twice before resolution, your definitions drift from your reality: simplify the rules, because two smart rules outperform eight clever ones at every team size. Revisit the setup quarterly; the small-team KPI guide has the two numbers — reassignment rate and unclaimed age — that say it's time.

Every model here works on the free tier, on three seats, today: routing lives in Settings, costs nothing extra, and travels with you if the desk later moves on-premise. Assignment is not a growth-stage feature — it is a five-minute setup that a two-person team benefits from on day one, and the two numbers it optimizes — reassignment rate and unclaimed age — are the same ones your customers feel at the other end of the chat. Get those small, and everything downstream, from response times to free-tier honesty, gets easier.