Conversation assignment rules are the instructions that decide who owns an incoming customer conversation: which person or team gets it, in what order, and what happens when that person is away, full or finished. Round robin is the best-known method, but it is only one of several, and it works properly only when it sits inside a wider set of rules about availability, workload, skills and response targets.
This guide explains each method in plain terms, shows how to write the rules so they hold up on a busy Monday, and covers how a conversation SLA policy connects to assignment. It is written for Australian service businesses that handle calls, chats and messages across more than one channel.
WattleKeep Customer Conversations MovingWattle brings calls, SMS and website chats into one inbox, where your team can review conversations and take over AI interactions.Explore WattleTL;DR
- An assignment rule answers three questions: who is eligible, who gets the next conversation, and what happens when nobody can take it.
- Round robin is fair on count, not on effort. Add availability and workload limits, or one person ends up with all the hard cases.
- Every conversation needs exactly one owner at any moment, and the SLA clock should follow that owner.
- Write one rule set for every channel (phone, web chat, WhatsApp, SMS), then adjust only the response targets per channel.
- Measure a baseline for at least 30 days before changing the rules, then change one thing at a time.
Table of Contents
- What conversation assignment rules actually do
- Round robin and the other selection methods
- Fair distribution across shifts and different workloads
- What happens when someone is away, full or hands off
- Writing a conversation SLA policy that matches your assignment
- Applying the same rules across phone, chat, email and messaging
- What to measure and when to change the rules
- Our perspective on assignment
- How Wattle fits into your assignment setup
- FAQ
- Sources
What conversation assignment rules actually do
A conversation arrives. Something has to decide where it goes. That decision can be manual (someone in the office glances at the inbox and says “I’ll take it”), semi-automatic (the conversation lands in a shared queue and people pick), or fully rule-driven (the system selects an owner).
A complete rule has four parts:
- Trigger. What starts the rule: a new conversation, a customer reply on a closed thread, a handover from an automated agent, or a manual transfer.
- Eligibility. Who may receive it: a named team, people with a particular skill, people on shift, people below their workload limit.
- Selection method. How one person is chosen from the eligible group: round robin, fewest open conversations, longest idle, or a fixed owner for a customer.
- Fallback. What happens when nobody is eligible: a overflow team, a manager, a voicemail-style callback task, or a holding message with a promised time.
Most teams write the first three and forget the fourth. The fallback is the part customers actually feel, because it covers evenings, lunch breaks, sick days and sudden spikes.
Round robin and the other selection methods
Round robin hands each new conversation to the next eligible person in a fixed rotation: Sam, then Priya, then Jo, then back to Sam. It is simple to explain and easy to audit, which is why it is the default in so many tools and why people search for it so often.
Its weakness is that it counts conversations, not effort. A five-minute booking change and a forty-minute complaint count the same. If Priya gets three complaints in a row, the rotation still says her turn is over.
Method How it picks the owner Works well for Watch out for Round robin Next person in a fixed order New sales leads, similar-length enquiries Ignores current load and who is away Least busy (balanced) Person with the fewest open conversations Mixed support work with uneven lengths Needs a sensible definition of “open” Skills-based Person or team tagged for the topic or language Licensed work, technical issues, multiple languages Skill tags go stale if nobody maintains them Priority-based Urgent items go to senior or on-call staff first Emergencies, high-value jobs Priority labels must be applied consistently Fixed owner Same person as the customer’s last conversation Account management, ongoing jobs Breaks when that person is away Manual claim from queue Staff pick from a shared list Small teams, specialist work Cherry-picking easy items, slow pickupIn practice you combine them. A common pattern is: filter by skill first, then by availability, then pick the person with the fewest open conversations, and use round robin only to break ties. That gives you fairness without ignoring the real work in front of each person.
Uneven splits
Sometimes equal is the wrong goal. A senior technician who also supervises might take half the share of a new starter, or a part-timer might take a share proportional to their hours. Weighted rotation handles this, and so does a workload limit set per person. If your tool only offers plain round robin, a workaround is to put a person in the rotation twice, though that is a blunt instrument and it hides the logic from the next person who reads the settings.
Fair distribution across shifts and different workloads
Fairness is where most assignment rules go wrong, because people work different hours and different speeds.
Tie eligibility to shifts, not to a login. If someone is logged in but on a break, they should not be eligible. The cleanest approach is a status that people can set (available, away, in wrap-up) plus a schedule that sets it automatically. Your business-hours rules should already define when the front desk is open, closed or on holiday, and our guide to business hours routing covers how to set up calls for each of those states.
Set a workload limit per person. A limit says: once you have this many open conversations, you stop receiving new ones until you close some. Choose the number by channel. A person can reasonably juggle several text chats at once but only one live call.
Allow for wrap-up time. After a call, people need a minute or two to write notes and update the job. If the next call arrives the second the last one ends, notes get skipped and the next conversation starts from nothing.
Track the long-term spread, not just the day. Round robin looks fair within a day and unfair across a month if one person is always on the late shift when the difficult after-hours cases arrive. Review the share of conversations, the average handling time and the number of escalations per person each month.
What happens when someone is away, full or hands off
This is the part to write down explicitly, because the default behaviour of most systems surprises people.
When an agent is away
Decide whether an away person is skipped by automatic assignment. Usually yes. Also decide what happens to conversations already assigned to them when they go away: stay with them, return to the team queue after a set time, or move to a named backup. For anything time-sensitive, a timed return to the queue is safer than leaving a customer waiting for someone’s lunch break to end.
When everyone is at their limit
Overflow needs a destination. Options include a wider team, a manager, a callback task with a promised time, or an automated holding reply that is honest about the wait. Do not let conversations sit unassigned with no clock running. An unassigned conversation is one that nobody is accountable for.
When a reply or handoff changes ownership
Two situations cause the most confusion:
- A teammate replies on a conversation owned by someone else. Decide whether replying makes you the owner. In some tools it does, silently, which scatters ownership. Make it an explicit action, or let the original owner stay owner with the replier as a collaborator.
- A customer replies after the conversation was closed. Decide whether it reopens for the previous owner (if they are available), goes back through assignment, or starts a new conversation.
Also note that different mechanisms in the same tool, such as an automated workflow and an assignment setting, can follow different rules about who is away or at their limit. Test both paths with the same scenario before you trust either.
When an automated agent hands over
When an AI front desk answers first, the handoff is itself an assignment trigger. The rule should say which conversations go to a person (a sensitive complaint, an urgent job, a customer who asks for a human), who receives them, and what context travels with them. We cover how to write and test these in our guide to call escalation rules. A good handoff passes the transcript, the customer’s details and the reason for the escalation, so the person never asks the customer to start again.
Writing a conversation SLA policy that matches your assignment
An SLA (service level agreement) policy sets the response and resolution targets you commit to. Assignment and SLA are two halves of one system: the SLA says how fast, assignment says who is responsible for that speed.
The three measures
- First response time: from the customer’s first message to a meaningful reply from a person or agent.
- Next response time: from a customer’s follow-up to your next update.
- Resolution time: from the start of the conversation to the issue being solved.
Set targets by priority and channel, not as one number. A burst pipe and a quote request should not share a clock. A phone call is answered live or not at all, while an SMS can reasonably wait a few minutes. Avoid copying a figure from a blog post. Base the first targets on what your team actually does today, then tighten them.
How assignment affects the clock
- Who is on the clock? The clock should show the current owner. When ownership changes, the clock keeps running, but the new owner sees how much time is left.
- Which hours count? If your targets apply during business hours only, the clock must pause overnight and on public holidays. Otherwise every Monday morning conversation starts already breached.
- What triggers an alert? Warn the owner before the target is missed, and alert a manager if it is missed. A breach with no follow-up is just a statistic.
- What about unassigned conversations? They should be on the clock too, with the alert going to whoever supervises the queue.
A worked example
Take a small electrical business with four office staff and an AI front desk answering first.
- Calls flagged as urgent (no power, burning smell) go straight to the on-call person with a target of an immediate transfer. If they do not answer, the call goes to the second on-call, then the office manager.
- Quote requests are booked by the AI front desk against the calendar. If a customer wants to discuss scope, it creates a callback task for the next available estimator with a first-response target of two business hours.
- Invoice questions go to the person with the accounts skill. Among the eligible staff, the one with the fewest open conversations gets it, with a resolution target of one business day.
- Anything unassigned for 15 minutes during business hours alerts the office manager.
That is five lines, and it covers eligibility, selection, fallback and the clock. The specific numbers are examples, so replace them with ones from your own history.
Applying the same rules across phone, chat, email and messaging
Customers do not think in channels. A person might phone, then follow up on WhatsApp, then send an SMS to confirm. If each channel has its own rules and its own queue, that one customer becomes three conversations with three owners.
To keep things consistent:
- Link every message to one customer record. Match by phone number or email so that the history appears together. Our guide to server-side CRM call logging explains how to avoid calls that land in the CRM unmatched.
- Use one eligibility and selection logic. Skills, availability and limits apply to everyone, whichever channel the message came through.
- Vary only the targets and limits. Live voice has a one-at-a-time limit and a seconds-level target. Chat and messaging allow more concurrency and longer waits. Email suits resolution-based targets.
- Keep a single inbox. When staff see all channels in one place, they can spot a customer who has contacted you twice and avoid duplicate work.
- Route by intent, not by door. A billing question should reach the billing person whether it arrived by phone or web chat. For a deeper look at this idea, see our piece on intelligent call routing.
What to measure and when to change the rules
Before you adjust anything, collect a baseline. One recommended approach for call-flow testing is to connect your call logs, interaction data and downstream outcomes, including whether callers reached the right agent or were misrouted, and to measure for at least 30 days before making targeted changes.
Useful measures for assignment rules:
- Share of conversations reaching the right person the first time
- Number of reassignments per conversation (more than one is a sign the rules are wrong)
- Time unassigned
- First response and resolution times against target, by priority and channel
- Spread of workload per person, over a month
- Customer satisfaction after closed conversations, where you collect it
Change one rule at a time, give it a couple of weeks, and compare with the baseline. If you change the limit, the skill tags and the rotation all at once, you will not know which helped.
Our perspective on assignment
We see assignment rules go wrong for the same reason again and again: they are written for the busy middle of the day and fall apart at the edges. The call at 5:40 pm, the message that arrives while the only qualified person is on a job site, the customer who rings back after their conversation was closed.
That is why we build Wattle as a front desk first and a routing tool second. An agent picks up every call, asks the questions that decide where the enquiry should go, books the work where it can, and passes the conversation to a person when it should not be automated. You decide the agent’s role, instructions and guardrails, and you control the handoff, so assignment starts with a clear reason and clear context rather than a ringing phone.
Because phone, web chat, WhatsApp and SMS all land in one inbox, one customer is one conversation, not three. We think that is the foundation of fair assignment: you cannot share work fairly if you cannot see it all in one place.
Christopher
How Wattle fits into your assignment setup
If your team is spending the day answering the same booking and enquiry calls, the first assignment rule to write is the one that stops those calls reaching people at all. Wattle answers calls and chats, qualifies the enquiry, books appointments against your connected calendar, and hands the conversations that need a person to your staff, with the transcript attached.
Plans are published as Starter at A$99 per month, Pro at A$499 per month and Max at A$999 per month, and you can see what each includes on the pricing page. To see how the agents, channels and handoff controls work together, browse the AI receptionist features.
FAQ
What is the difference between round robin and load-balanced assignment?
Round robin gives each new conversation to the next person in a fixed order, regardless of how busy they are. Load-balanced assignment gives it to the eligible person with the fewest open conversations. Load balancing copes better when conversations vary in length, while round robin suits similar, short enquiries such as new leads.
Should away staff receive conversations?
No, not by automatic assignment. Skip anyone marked away or on a break, and decide in advance whether their existing conversations stay with them or return to the queue after a set time. For urgent work, a timed return stops a customer waiting for someone to come back.
Who owns the SLA when a conversation is transferred?
The current owner is responsible for the next step, but the clock should keep running from the original start so the customer’s total wait is visible. Show the new owner how much time remains, and alert a manager if the target is about to be missed.
Can I use the same rules for phone and messaging?
Yes. Use the same eligibility, skills and fallback logic everywhere, and vary only the concurrency limits and response targets. A person can handle one live call at a time but several text conversations, so the limits and targets should differ by channel.