Customers can reach a brand through live chat, self-service portals, social media, and apps, but not every channel suits every problem, and many customers still prefer to pick up the phone.
When lines are busy, companies often try to redirect callers with messages like, “Did you know you can solve your issue on our website?” That helps some callers. Others can’t or don’t want to go online, which is why they called in the first place.
Done well, call deflection shortens queues and saves time for your customers and your team. Done badly, it moves frustrated callers somewhere else and leaves the problem unsolved. The difference comes down to one question: did the customer’s problem actually get solved? Here’s how to get it right.
What is call deflection?
Call deflection is a strategy for reducing inbound call volume by guiding customers to other support channels, such as IVR self-service, FAQs, chatbots, or live chat, so their issue can be handled without a live agent. The goal is to resolve the issue in that channel, not just to keep the caller out of the queue.
That distinction matters because three metrics often get used interchangeably. Deflection counts calls moved away from agents, containment counts self-service sessions that finish without a person stepping in, and resolution counts callers whose problem was actually solved. Only resolution tells you whether the customer got what they came for.
| Metric | What it counts | What it can’t tell you |
|---|---|---|
| Deflection rate | Contacts moved off the agent queue ÷ total contact attempts × 100 | Whether the new channel solved anything |
| Containment rate | Self-service sessions completed without escalation | Whether the caller hung up frustrated or called back later |
| Verified resolution | Actions completed in the system of record and confirmed by the customer | What it cost to get there, so pair it with cost per contact |
How does call deflection work?
Call deflection uses a mix of channels and tools to handle customer questions outside the traditional contact center queue. The strongest strategies match each type of request to a channel that can actually solve it. These are the most common.
IVR
Touch-tone menus (“press 1 for billing”) and keyword-driven IVRs (“tell us in one word why you’re calling”) are usually the first layer of IVR deflection. They route callers to a solution based on predefined prompts. A caller with a billing question, for example, may hear an option to pay online instead of waiting for an agent.
Email lets customers ask for help on their own schedule. It suits non-urgent questions and issues that need a detailed answer, like starting a product return.
Live chat and messaging
Live chat and messaging connect customers with support agents in real time without a phone call. A customer tracking an order, for instance, can get an update in chat instead of waiting on hold.
Self-service tools (FAQs and chatbots)
FAQ pages and knowledge bases give customers instant answers to common questions. Chatbots guide users through troubleshooting or answer simple questions from pre-programmed responses, and they work best for requests like “How do I reset my password?” or “What are your store hours?”
What are the benefits of call deflection?
When it’s done well, call deflection makes life easier for customers and agents. Fewer people wait on hold, and your team spends more time with the customers who need them most. The main benefits include:
- Improved efficiency: fewer routine calls free your agents to focus on complex conversations.
- Lower cost: a self-service contact costs a fraction of an assisted one (see the benchmarks below).
- Customer satisfaction: customers who solve their issue without waiting on hold have a better experience.
What are the challenges of call deflection?
Poorly executed deflection backfires. Sending customers to a channel that can’t solve their problem just moves the problem around, and it often brings them back to the phone more frustrated than before.
Gartner surveyed 5,728 customers and found that 73% of customers use self-service at some point in their journey, but only 14% of service issues are fully resolved there. Even issues customers described as “very simple” were resolved in self-service only 36% of the time. A separate Gartner study of 1,492 customers found 62% of channel transitions felt high-effort, though 74% of customers who had an easy handoff said they’d try self-service again.
Voice self-service has its own history of frustration. Traditional IVRs often misunderstand callers, make them repeat themselves, and can’t handle clarifying questions, which is how callers end up shouting “Agent!” to escape the loop. Once customers get stuck like that, they stop trusting the system, and call abandonment rates rise.
What does it cost when a call moves between channels?
Every channel a customer passes through also adds cost. Gartner’s February 2024 cost benchmarks , and the journey data CX Dive reported , show that resolution costs rise sharply as a journey spreads across channels.
| Measure | Cost (USD) |
|---|---|
| Median cost per self-service contact | $1.84 |
| Median cost per assisted contact | $13.50 |
| Average cost to resolve in one channel | $9.10 |
| Average cost to resolve across three channels | $26.04 |
A customer who moves from self-service to a text link to an agent costs nearly three times as much to help as one whose issue gets solved in the conversation they started.
Why is resolution the number to track?
Resolution is the only measure that tells you whether the customer's problem was solved. Deflection and containment show where a call went and whether it stayed there, which helps you manage workload and cost. A caller can still be deflected and contained and hang up with nothing fixed.
What you measure shapes what your team builds. When the target is deflection, the quickest way to hit it is to make a person harder to reach, and the costs from the sections above follow: repeat calls, journeys that spread across channels, and customers who stop trusting the system. When the target is resolution, every change has to make it easier for the customer to get the work done.
PolyAI measures its dialog agents on verified resolution, meaning the action is in the system of record, and the customer has confirmed it.
How does a dialog agent resolve calls?
Self-service channels like IVRs, chatbots, and FAQ pages can take some volume off the queue, but as the Gartner figures show, most issues aren’t fully resolved there. When self-service falls short, customers look for someone who can help, and the phone call becomes the moment where resolution matters most. That’s where a dialog agent does its work.
A dialog agent understands what the caller says in their own words and acts directly in your systems, then confirms the result before the call ends. The booking gets made, the payment goes through, or the appointment moves, and the record updates while the caller is still on the line.
PolyAI’s Dialog-RSN-1 handles turn-taking, speech recognition, function calling, and the response in one audio-native model. It hears the caller’s tone, their hesitation before an answer, and the noise around them, and it responds in under 300 milliseconds.
Fogo de Chão’s dialog agent answers 100% of guest calls and completes 88% of bookings, with 95% guest satisfaction. When outages hit, PG&E customers calling for urgent information get an answer immediately, and CSAT on outage calls rose 22%. PolyAI’s largest deployments do the work of more than a thousand full-time employees, and brands including Marriott, FedEx, and UniCredit run PolyAI in production.
When should a call go to a person?
There’s no universal list of calls that must start with a person. Some organizations send complaints, bereavement, fraud disputes, or medical emergencies straight to their team; others let a dialog agent take the first turn and bring a person in when the conversation needs one. That decision belongs to each business and each use case. What matters is that the handoff can happen at any moment, and the caller reaches the right person quickly, with the full conversation so far already in front of them.
SQM’s 2024 benchmark across more than 500 North American contact centers shows why the hardest calls deserve your team’s time. Complaints reach first contact resolution 48% of the time, compared with 73% for general inquiries. A dialog agent that takes on routine work gives your people more room for the conversations where their judgment makes the difference.
How do you keep improving resolution over time?
Improvement works best when it’s continuous and stays under your control. PolyAI’s Wren reviews every conversation, drafts a fix, and proves it with a test, and when it matters most, a person makes the final call before anything goes live. Wren filed more than 6,000 recommendations against production conversations in 30 days. Golden Nugget saw bookings rise 33%, resolution rise 10%, and conversion rise 12% within 24 hours of its first approved changes.
Did the work get done?
Every call starts with a customer who needs something done, and resolution is the only number that tells you whether they got it. Measure it first, and let it guide which calls a dialog agent handles, where your team steps in, and what you improve next. If you’d like to see what resolution looks like against your own call data, the PolyAI team is happy to help.
The conversation is only finished when the work is.
Frequently asked questions
-
Call deflection moves a caller to a channel that can solve their issue and keeps a person within reach. Call avoidance steers callers away from help without solving anything, which usually leads to repeat calls and lost customers.
-
Call forwarding sends a call to another number, usually a different agent, team, or location. Call deflection redirects the customer to another channel, like self-service or chat, before they reach an agent.
-
Divide the number of contacts moved off the agent queue by total contact attempts, then multiply by 100. Decide up front what counts as a contact attempt, since teams define it differently, and track resolution alongside it so a high deflection rate can’t hide problems that were never solved.
-
Start with verified resolution, then track CSAT or NPS, FCR, AHT, and cost per contact, with deflection and containment rates as supporting metrics. Resolution tells you whether the work got done, and the rest tell you how efficiently it happened.
-
A dialog agent takes on routine, high-volume work like bookings, rescheduling, and account changes, which frees your team for complaints, disputes, and calls that need a person. With average agent attrition at 52% in 2023, according to Deloitte’s 2024 Global Contact Center Survey , easing routine volume also helps you keep the people you’ve trained.
