Call center coaching is a short one-on-one, usually weekly, in which a team leader and an agent listen to a few of the agent's own calls, agree on one thing to do differently, and check on later calls whether it changed. Quality assurance scores calls against a standard. Coaching takes what the scores and the dashboard show and turns it into a call that goes differently next week.
This guide gives you a routine that takes about 45 minutes of a team leader's time per agent each week, a table for tracing a number on the dashboard back to the behavior behind it, a 20-minute session plan, and a coaching form you can copy into a spreadsheet or your QA tool.
What is call center coaching, and how is it different from QA?
Call center coaching is regular one-to-one work on how an agent handles calls. You choose one behavior, find it on the agent's recordings, agree on the change, practice it, and check it on new calls. QA measures a sample of calls against a written standard. Coaching starts from what QA and the numbers find, and it has worked only when the behavior changes.
McKinsey's operations practice found that supervisors often blur coaching with incidental feedback and reminders about process, and wrote that "effective coaching should look beyond the here and now to focus on the skills, attitudes, and behaviors that will improve agent performance over time" (Smarter call-center coaching for the digital world, 2018). Reminding an agent to use the new greeting is feedback. Working out why the same agent puts callers on hold three times on every billing call, and fixing that, is coaching.
How much coaching an agent gets matters too. In a study of call center agents whose calls were handed out at random by the routing system, the amount of coaching an agent received each month predicted objective performance improvements over time (Liu and Batt, How supervisors influence performance, Personnel Psychology, 2010).
How often should you coach call center agents?
Weekly, in sessions of 15 to 30 minutes, with a few minutes of listening beside each agent on live calls in between. Gallup's research on manager conversations found that 15 to 30 minutes is enough for a meaningful conversation if it happens often, and that regular conversations of that length have more impact than 30 to 60 minute ones (A Great Manager's Most Important Habit, updated February 2026).
The same Gallup piece reports that 80 percent of employees who say they received meaningful feedback in the past week are fully engaged. A monthly review leaves most agents without that in most weeks.
Here is what a weekly routine costs a team leader with 12 agents. The inputs are an example to show the arithmetic, not a benchmark:
| Task | Per agent each week | Team of 12 each week |
|---|---|---|
| Choosing and listening to 2 or 3 calls | 15 minutes | 3 hours |
| The coaching session | 20 minutes | 4 hours |
| Checking the change on new calls | 10 minutes | 2 hours |
| Team leader total | 45 minutes | 9 hours |
| Agent time off the phones | 20 minutes | 4 agent hours |
Nine hours is a little under a quarter of a 40-hour week. McKinsey's guidance is that "in most contact centers, supervisors should aim to spend at least half their time coaching agents on the front line," so treat the weekly sessions as the floor and give the rest of that time to listening beside agents. For each agent, 20 minutes is under 1 percent of a 40-hour week. It still counts as planned shrinkage, so book it in the half hours that have spare agents (call center shrinkage).
A weekly call center coaching routine
Plan on Monday, coach midweek, check on Friday. Each agent works on one behavior a week, picked from the numbers and confirmed on recordings. The Friday check decides whether that behavior stays on the agent's form for another week or the agent moves on to the next one.
| When | What the team leader does | Time |
|---|---|---|
| Monday | Pull last week's numbers by agent and call type, pick one behavior per agent with the table below, and choose 2 or 3 calls that show it | 15 minutes per agent |
| Tuesday to Thursday | Run the 20-minute sessions, booked in quiet half hours | 20 minutes per agent |
| Every day | Listen beside one or two agents on live calls, on the behavior each is working on | A few minutes per agent |
| Friday | Listen to 2 or 3 new calls of the same type per agent, mark the change as held, partly held or not yet, and set next week's behavior | 10 minutes per agent |
| Once a month | A manager sits in on one session per team leader and coaches the team leader afterward | 20 minutes per team leader |
Don't drop the monthly row. McKinsey recommends that managers sit in on coaching sessions from time to time, and quotes a manager who did it after watching a supervisor with 10 years of experience: "I had no idea there was that big a difference between what was actually happening and what I thought was happening in the coaching sessions."
How do you decide what to coach?
Start from a number, then find the behavior on recordings before you say anything to the agent. McKinsey warns against "coaching to a number," singling out one metric and urging the agent to improve it "without any effort to identify the underlying behavior driving the metric," which it calls "ineffective and often counterproductive to employee engagement." A handle time gap tells you where to listen, and the recordings show what to coach, such as three holds on every billing call.
Use this table to get from the signal to what you'll coach. Each row is a place to start listening, and the recordings decide.
| The signal | What to listen for | What to coach |
|---|---|---|
| Handle time well above the team on one call type, with long holds | The agent putting callers on hold to find an answer or ask a colleague | Where the answer lives: walk the lookup together, and fix the article if the article is the problem (average handle time) |
| Long after-call work | Notes written from scratch after the caller hangs up, or a long hunt for the right code | A note pattern and the codes that fit the call type (after-call work, disposition codes) |
| Many transfers on a call type the agent is allowed to resolve | A transfer before the caller has finished explaining | What the agent can resolve alone, and the questions to ask before transferring |
| Callers calling back about the same issue within a few days | A call that ends without a stated next step or date | Closing the call: what happens next, when, and what the caller should do if it doesn't (repeat contact rate) |
| Low survey scores on calls that look fine on time | Talking over the caller, a rushed greeting, a script read flat | Pacing and acknowledgment: play 30 seconds back and let the agent hear it (call center scripts) |
| Escalations or requests for a supervisor | The moment the caller's tone changed and what the agent said next | One move for that moment: acknowledge, say what you can do, give a time |
| Low schedule adherence | Nothing on the calls: late returns from breaks, unplanned off-phone time | A conversation about the schedule, not a call review (schedule adherence) |
The last row isn't about call handling at all. McKinsey calls schedule adherence "a will-driven metric," as opposed to a skill one, and a call review won't change it. And when the same knowledge gap turns up across several agents, fix the article or the process once instead of coaching it agent by agent.
Which calls should you use in a coaching session?
Use two or three calls of one call type that show the behavior, plus one call where the agent handled it well. McKinsey notes that coaches often work from "four to six calls out of the five or six hundred the agent conducts every month." That is far too few to score an agent on, but it is enough to coach one behavior when the calls were chosen for that reason.
Keeping to one call type means the agent can't wave a clip off as an odd call. Write down the call ID and the minute where the behavior shows, so the session plays 30 seconds instead of eight minutes. If you need to compare agents or report a quality score, the sample has to be much larger, and the call center QA guide works through how much larger.
A call from your strongest agent on the same call type is worth adding. In an NBER study of 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14 percent on average and by 34 percent for novice and low-skilled workers, and the authors found suggestive evidence that it worked by passing along the best practices of more able workers (Generative AI at Work, Brynjolfsson, Li and Raymond). Playing a newer agent the way a strong colleague handles the same moment is a manual version of that.
A call center coaching form you can copy
Use one form per agent per week, with one behavior on it. Fill in the first six fields before the session, the next three during it, and the last one at the Friday check. The example column is made up to show the level of detail.
| Field | What to write | Example |
|---|---|---|
| Call type | The one call type this week's behavior applies to | Billing plan change |
| Signal | The number that pointed here, with the team figure beside it | Hold time per call 2:10, team average 0:50 |
| Behavior | What the agent does, in words you could hear on a recording | Puts the caller on hold to find the proration rule |
| Evidence | Call IDs and the minute where it happens | Two calls, at 1:40 and 3:05 |
| What went well | One call and moment the agent handled well, raised first | Clear recap of the new plan before ending the call |
| Agent's read | What the agent noticed when listening, in their words | "I never know which screen has the proration rule" |
| Agreed change | One sentence, in the agent's words | "Open the plan screen before I ask which plan they want" |
| Practice | The step or line rehearsed in the session | Walked the lookup twice on a test account |
| Check | Which calls, how many, by when | Three billing plan changes, by Friday |
| Result | Held, partly held or not yet, and next week's behavior | Held on all three. Next: state the next step before ending the call |
When the agent's read points at a tool or an article, as it does in the example, send it to whoever owns the knowledge base too. Other agents may be holding callers on the same screen.
How to run a 20-minute coaching session
Open with what went well, play the clip and let the agent speak first, agree on one change in the agent's words, practice it, and set the check. Gallup found that recognition for recent work is the top characteristic of conversations employees call meaningful, and that their weaknesses are the one topic they see as less meaningful.
| Minutes | Step | What to do |
|---|---|---|
| 0 to 3 | What went well | Play or describe the good moment and say exactly what worked |
| 3 to 10 | The clip | Play the 30 to 60 seconds that show the behavior, then ask what the agent heard before you say anything |
| 10 to 14 | One change | Agree on the change and write it on the form in the agent's words |
| 14 to 18 | Practice | Rehearse it twice, on a test account or as a short role play |
| 18 to 20 | The check | Agree which calls you'll review, how many and by when |
If the agent disputes what the clip shows, play the second call instead of arguing about the first. Keep the session to one behavior, because the Friday check can only test one.
How do you know whether coaching is working?
Check three things on three different clocks. On Friday, see whether the agreed change shows up on new calls. Over the next several weeks, see whether the number that pointed to it moves on that call type. Over months, see whether the gap between your strongest and weakest agents narrows.
McKinsey estimated that an average 500-agent call center in North America puts around $2 million of staff time into coaching each year, and found that "a lot of that coaching doesn't seem to be working." For agents more than six months in the role, handle time, first-contact resolution and satisfaction should improve over time and the differences between agents should shrink. A gap of much more than 30 to 40 percentage points between the top and bottom quartiles, it wrote, "suggests weaknesses in the coaching regime."
Track the coaching itself as well as the results. McKinsey found that less coaching happens than companies expect, because team leaders lose their slots to meetings and admin. One company it studied found that the time managers spent coaching on the floor varied by a factor of six, and teams whose coaches spent more than 60 percent of their time on the floor had twice the average retention. Log the minutes each agent was actually coached, not the sessions booked, and read them next to your attrition rate and the rest of your metrics and operations numbers.
What changes in coaching when AI agents take some call types?
The calls left for people get harder, so coaching moves away from scripts and speed and toward judgment, problem solving and picking up a call someone else started. McKinsey described the same shift as digital channels took simple requests: agents spend "much less of their time handling simple, transactional calls," and 94 percent of the more than 50 senior customer care executives it surveyed expected the skill demands on their agents to rise.
- Take the call types an AI agent now handles off the coaching forms, and spend that time on the ones still with your team.
- Coach the pickup of a transferred call. The caller has already explained once, and a greeting that restarts the script makes them do it again. Coach a line that asks where things stand before anything else.
- Read the AI agent's calls before Monday's planning. Its summaries show what callers on that call type asked for that week, which tells you what your people will hear next.
- Fix the AI agent through its instructions, not coaching. A change to its instructions applies to every later call, and its calls can be scored on the same scorecard as your team's (call center QA).
For where AI agents fit in a center that's already running, see the guide to adding AI agents to a call center.
How Telvana's call records fit a coaching week
Every call an AI agent from Telvana takes is recorded, transcribed, summarized, tagged by outcome and given a sentiment, and the calls can be listed by agent, date range and sentiment (call results). A team leader can read a week of summaries for one call type during Monday's planning and see what callers asked for before the sessions start.
Two limits matter for coaching. The records cover the calls Telvana's agents take, not the calls your people take on your own phone system, so your recording and QA tools still supply those. And transfers are cold: the person who picks up gets the caller without a summary, which is why the pickup line above is worth a session. Our team builds each agent from your scripts and transfer rules and tests it on your call types before it takes a live call. Pricing starts at $399 a month based on call volume, plus a one-time setup fee (pricing).
Frequently asked questions
What is coaching in a call center?
It is regular one-to-one work between a team leader and an agent on how the agent handles calls. The two listen to a few real recordings, agree on one behavior to change, practice it, and check it on later calls. It differs from feedback, which is a passing correction, and from QA, which scores a sample of calls against a standard.
How often should call center agents be coached?
Weekly is a sound default: one session of 15 to 30 minutes plus a few minutes of listening beside the agent on live calls. Gallup found that 15 to 30 minute conversations held regularly have more impact than 30 to 60 minute ones. New agents and agents working on a hard call type may need more.
How long should a call center coaching session be?
About 20 minutes is enough when the calls are chosen and timestamped beforehand. Spend the first few minutes on what went well, most of the time on one clip and one change, and the last minutes practicing it and agreeing how you'll check it.
What is the difference between coaching and QA in a call center?
QA scores a sample of calls against a written standard and tells you where quality falls short. Coaching takes one finding, traces it to a behavior on the agent's own calls, and changes it. QA needs large samples to compare agents fairly; coaching needs only a few calls chosen for a reason. The QA guide covers the scoring side.
What should a call center coaching form include?
The call type, the number that pointed to the issue, the behavior in plain words, call IDs with timestamps, one thing that went well, the agent's own read, the agreed change in the agent's words, what was practiced, how it will be checked, and the result. The form above has an example of each.
What is the 80/20 rule in a call center?
It usually means a service level target of 80 percent of calls answered within 20 seconds. It is a common convention, and each center sets its own target. The service level guide explains where 80/20 comes from and how to calculate it.
Give your coaching time to the calls that need people
Book a demo and bring your highest-volume call types. We'll show you which of them an AI agent could take, so team leaders can coach on the calls that stay with your people, and you can hear a Telvana agent handle one.