Call routing
In one sentence
Call routing is the logic that sits between a call arriving and someone answering it, deciding which queue, agent, department or automated system the call should reach, and an AI voice agent typically joins that logic as one of its destinations rather than replacing it.
Not to be confused with IVR.
Definition
Call routing is the decision about where an incoming call goes: which queue, which agent, which department, or which automated system.
It is a mature discipline with decades of practice behind it, and an AI voice agent almost always plugs into it rather than replacing it.
Routing decides the destination. It is the layer between the call landing and a human or a machine picking it up, and every contact center runs one whether or not anybody calls it by name.
The routing strategies in common use
- Skills-based. Match the caller need to an agent who has the relevant capability.
- Round-robin and longest-idle. Distribute evenly across whoever is available.
- Priority-based. Certain callers or issue types jump the queue.
- Data-directed. Route on what is already known about the caller, such as account value or an open case.
- Time-based. Different destinations by hour, day or holiday.
- Percentage allocation. Split traffic, most often to run a test.
The four places a voice agent can sit
- As a destination. Calls route to the agent instead of a queue.
- As a router. The agent determines intent and routes accordingly, which is conversational IVR.
- As an overflow. Calls route to the agent when queues cross a threshold or fall outside hours.
- As a filter. The agent resolves what it can and passes the remainder onward with context.
- Overflow and filter are the easiest positions to sell into an established contact center, because they add capacity without displacing anything.
Why routing accuracy matters more than it looks
- A misrouted call spends two agents' time and the caller's patience.
- Transfers between queues are a leading source of dissatisfaction.
- Routing on the caller's own description of the problem is generally more accurate than routing on a menu selection, which is the core argument for conversational routing.
The context problem at the handoff
- Routing decides where a call goes. Whether the destination receives any context is a separate question, and frequently the answer is no.
- That gap is where most of the experience damage happens, and it is covered in the warm and cold transfer entries.
The website equivalent
- A website has routing too, in the form of navigation, and it carries the same flaw: it asks the visitor to work out which category their question belongs to.
- On-site voice replaces category selection with a description, which is the same improvement conversational routing makes on the phone.
Common misconception
That deploying a voice agent replaces routing. In most established deployments the agent becomes a participant in existing routing rather than a replacement for it, which is both easier to sell and easier to implement.
Why it matters commercially
Routing is the seam where a voice agent meets an existing contact center. The ability to slot into the routing already running, rather than demand a replacement for it, is what decides how readily an enterprise will deploy, so leading with the overflow or filter position closes more than leading with a rebuild.
In voice specifically
Routing on a spoken description beats routing on a keypad selection, and that difference is specific to voice. A menu forces the caller to translate their problem into one of a handful of numbered categories before the routing logic ever sees it. Let them describe the problem instead and the routing decision is made on what they actually said, not on the box they guessed at.
Where AsqVox fits
The website analogue of routing is navigation, and voice navigation addresses the same problem: the visitor describes what they want rather than deducing which menu contains it. AsqVox runs in the browser, so there is no telephony routing layer to integrate, only the same describe-instead-of-choose improvement applied to a site.
Visual
Four places a voice agent can sit
The agent joins existing routing. Overflow and filter add capacity without displacing anything.
Statistics
Every figure carries its source and year. Vendor numbers are labelled as vendor numbers, and where no reliable figure exists this page says so rather than borrowing one.
Containment sits at roughly 70 to 80 percent in mature deployments, 40 to 55 percent on average, and under 35 percent for rule-based systems. Conversational routing tends to improve routing accuracy without moving containment much, because routing and resolution are different operations.
70 to 80% mature, 40 to 55% averageindustry rangeIndustry-reported containment ranges, 2026 - Reported ranges rather than audited figures. The paired point about conversational routing is qualitative: better routing accuracy does not by itself lift the share of calls resolved without a human.
A human-handled call costs roughly USD 7 to USD 12 against roughly USD 0.40 for an agent-handled call. A misrouted call consumes the human cost twice.
USD 7 to 12 vs USD 0.40industry rangeWidely cited industry cost range, 2026 - A widely quoted range, not a metered figure, and it swings with country, vertical and complexity. The useful part is the order of magnitude between a human call and an automated one, and that a misroute pays the human rate on both ends.
Gartner forecast that by 2026 conversational AI in contact centers would reduce agent labor costs by USD 80 billion, with one in ten agent interactions automated by 2026.
USD 80 billion, 1 in 10 by 2026analyst forecastGartner press release, attributed to VP analyst Daniel O'Connell, 2022 - From the 31 August 2022 press release. Cite it as a dated analyst forecast, not as an outcome, and note the one-in-ten figure implies substantial ongoing human involvement rather than full automation.
There is no published cross-industry benchmark for routing accuracy, nor for the cost of misrouting, even though both are measurable inside any given deployment.
-no reliable figureMeasure them on your own traffic. Any single industry-wide accuracy or misroute-cost figure offered without a named sample is invented, because there is no study to draw one from.
There is no published benchmark quantifying how much conversational routing improves routing accuracy against menu-based routing.
-no reliable figureThe direction of the effect is well argued and the size is not measured in public. State that description-based routing is more accurate than menu selection and stop short of attaching a percentage.
Examples
In practice
An insurer deploys a voice agent as overflow, taking calls only when the queue wait crosses a threshold. Nothing about the existing routing changes, no numbers move, and the incumbent contact center platform stays in place, so internal resistance is minimal because nothing is displaced. Once the agent has demonstrated its performance, its role expands to filtering the highest-volume informational queries and passing the rest onward with context.
The everyday version
Call routing is deciding where a call goes when it arrives. A voice agent can be one of the places calls go, or it can be the thing that works out where they should go, or it can pick up the ones nobody else can get to. The last one is usually the easiest place to start.
Usage
Who says it
- Contact center operations teams and telephony engineers, fluently.
- It appears in every contact center platform and throughout migration planning.
Where it turns up
- In an RFP it sits next to routing strategies, queue integration, overflow behavior, out-of-hours handling, and context passing at transfer.
- The context-passing clause is the one that quietly decides whether escalation works, and it is frequently left unspecified.
Common misuse
- Assuming a voice agent replaces routing rather than joining it.
- Treating routing accuracy and resolution as the same improvement, when better routing does not by itself lift containment.
- Specifying routing without specifying whether context travels with the call.
Questions people ask
What is call routing?
Call routing is the logic that decides where an incoming call goes: which queue, which agent, which department, or which automated system. It sits between the call arriving and someone answering, and it runs on strategies such as skills-based, priority, data-directed and time-based routing. Every contact center has one whether or not anybody names it.
Does a voice agent replace call routing?
In most established deployments, no. The agent becomes a participant in the existing routing rather than a replacement for it. It can sit as a destination, as a router determining intent, as an overflow catching calls the queues cannot, or as a filter resolving what it can and passing the rest onward. Joining existing routing is both easier to sell and easier to implement than replacing it.
What is the easiest way to introduce a voice agent into an existing contact center?
The overflow and filter positions. As overflow the agent takes calls only when queues cross a threshold or fall outside hours; as a filter it resolves the queries it can and passes the remainder onward with context. Both add capacity without displacing anything, which keeps internal resistance low because nothing that already works is being taken away.
Why is routing on a spoken description more accurate than a phone menu?
A menu forces the caller to translate their problem into one of a few numbered categories before the routing logic ever sees it, and the guess is often wrong. Routing on the caller's own description makes the decision on what they actually said. That is the core argument for conversational routing, and it is the same improvement on-site voice makes over website navigation.
Last reviewed 4 August 2026. Written and reviewed by Dhruv Dholakia, founder of AsqVox.