I could be the nicest person in the world, but when the interaction drops into my headset, the customer is already yelling.
- They have been through the website FAQ, IVR, agentic voice AI, and now I am the fourth layer they are facing.
- They are frustrated, angry, and they want the company, meaning me since I answered the phone, to know how angry they are.
That is what agents are facing now. That is what makes emotional intelligence, de-escalation skills, and the ability for agents to emotionally reset after a call crucial for both customer retention and agent retention.
The Broken Training Model
With agentic AI voice and chat, every contact that moves to a human agent is an escalation. Those require more advanced skills and better training. However, many contact center training models were built in the 1990s and early 2000s. They desperately require updating for the age of agentic AI.
New hire agent training used to be about training people to handle the most common interactions, such as password resets and account balance inquiries.
So, after the new agents graduate, they can handle 80% of their interactions by themselves. Then with additional coaching, they can learn the rest on the job and handle rarer or more advanced interactions.
The old training goal was bare minimum at graduation, competency within six months, and mastery within one year.
AI throws all of that out the window.
Since AI is deflecting an increasing percentage of straightforward interactions, harder ones are coming to the agents. So, what might have been a once a week or once a day difficult interaction now becomes several times a day.
You cannot just train agents to 80% anymore. Instead, new hire training is going to have to cover:
- Emotionally intelligent customer service skills.
- Negotiation skills.
- How to deal with complexity.
- How to take escalated “handoffs” from AI, review the details AI has already gathered (to avoid making customers repeat themselves), and then start their own conversations with the customers.
Otherwise, when new hire training graduates hit the queues, they will feel like they were never properly trained in the first place because most interactions are those they were not trained to handle.
...many contact center training models were built in the 1990s and early 2000s. They desperately require updating for the age of agentic AI...
That frustrated feeling can breed resentment, a loss of confidence in the organization, and a feeling of “this is not what I signed up for.” These then lead to increased new hire attrition, further stressing your service levels.
Watch out for this by monitoring:
- Has the percentage of new hires quitting in the first two or three months increased versus pre-AI?
- Is your turnover rate higher at the six-month mark than it was?
Those are indicators that agents feel they were never properly trained in the first place.
Revamp Your New Hire Training
When I was the training manager for a national telco’s 900-agent contact center, one of the challenges was how to train new hires to do their jobs without making training take forever.
We had a six-week training program followed by two weeks of nesting, where new hires could sit together, take interactions, and get immediate help from veteran coaches.
However, feedback from past new hire classes showed they could benefit from additional customer service training.
We also needed to slow down the course somewhat, to give new hires a chance to digest the information without becoming overwhelmed.
There was just too much material compressed into too little time. This resulted in agents who could pass each test, but who could not necessarily apply that knowledge immediately to actual customer situations.
The course also put too much pressure on the nesting coaches. They ended up doing remedial training on concepts new hires did not have time to fully understand within the tight course timeframe.
That gap between passing a test in training and applying it under pressure is now the daily reality, as new hires face escalations from agentic AI as soon as they hit the contact center queues.
Yet every time I proposed making training longer, I got pushback from senior leaders. It was more important, they told me, to get bodies into the queue than to train for an extra week.
Their argument was training time costs money because new hires are in the classroom instead of being productive on the floor. But with today’s AI deflection making every human agent interaction an escalation, having better trained new hires is not a luxury. It is an essential.
If you do not do that, you are going to have to spend more time having a senior agent, team leader, or coach help agents deal with tough one-off interactions.
As I noted earlier, under the old new hire training paradigm, only the most common 80% of interaction types were covered in training. But if agentic AI is containing more of those, that number drops to perhaps 40% of a human agent’s daily interactions.
Based on this logic, the remaining 20% of rare interactions triples to 60% of a human agent’s interactions today. That means three times the volume of weird, complex, and emotional interactions. And most new hires are not receiving training for that.
The obvious solution is to increase the length of new hire training to accommodate more customer service and negotiation skills training, as well as training on how to solve more complex issues.
Of course, the key word is “productive.” If new hires cannot handle a heavy dose of agentic AI escalated interactions today, they are not productive. Instead, lack of training on tough issues makes them a potential liability. They could give out wrong information, credits, or aggravate a customer into leaving.
There are ways to make new hire training more effective though, without adding weeks to the course.
First, and ironically, AI is one way to drive training efficiency. Trainers should work with the QA (quality assurance) team and use AI-driven speech analytics to find the top interaction drivers escalated to human agents. As a trainer:
- Is your new hire course even designed around interaction drivers?
- Can you name the most common interactions AI escalates to your human agents?
- Is there an established workflow between your training team, QA team, and operations management to keep that curriculum current?
You need those feedback loops to make curriculum changes and keep your new hire training aligned with what is happening on the floor.
A second way to increase learning and save time is to make your training scenario based. For example, instead of just showing new hires how to add an address in your CRM, frame your training with the scenario of a customer calling in to change their address.
Scenario-based training is faster, more meaningful, and has better learner retention, because human agents first learn WHY they need to do something, then HOW to do it. That prepares them to handle actual customer scenarios versus learning facts or processes in a vacuum.
Add Training for Veteran Agents
It is not just new hires that need these advanced skills. Your existing veteran human agents may also need training on:
- Customer service skills.
- Emotional intelligence, empathy, and negotiation skills.
- How to solve more complex problems.
- The same training new hires receive about how to take escalated “handoffs” from AI and begin their conversations with customers.
The way to look at it is that the unusual, complex, or emotional situations that may have happened once or twice a day may now be happening a half dozen to a dozen times a day. This can increase your existing agents’ stress level.
Have you had an existing agent quit because they feel customer interactions are much harder than they used to be?
They are feeling pressure now from angrier customers and more complicated issues escalated to them by AI. That forces the agent(s) to cope with tougher interactions and more of them. They may have never been trained to do this.
Bottom line: it is more cost-effective to train existing agents on the skills they need to handle what agentic AI cannot do.
A recent Verint survey found “46% of agents aged 18–34 were likely to leave within six months.” To help retain agents instead, provide additional training on how to deal with these escalations, rather than train brand new hires on everything.
Back to “productivity.” Which is more cost-effective, no, or more additional training?
Consider this: McKinsey and Company research puts the cost of replacing a contact center agent at $10,000 to $20,000, a figure still widely cited across the industry today.
Bottom line: it is more cost-effective to train existing agents on the skills they need to handle what agentic AI cannot do.
Check the Effectiveness of Your Training
For new hires, check with the post-training “nesting team” to see if they are having to do more work now than they did before to fill in new hire skills gaps.
Also check with your QA team to see if the kinds of interactions they are scoring agents on match what is coming to human agents. The point is that new hire human agents need to know more to handle the escalations that are coming to them.
For both new hires and existing agents, one way to check for rising interaction complexity is to look at average talk time rather than average handle time (AHT).
The reason I mention talk time as opposed to AHT is AI agent assist, automatic note-taking, and faster backend processing could reduce after-call work (ACW). Reduction in ACW can mask an increase in talk time, because AHT may look the same or even shorter, even as your actual talk time is rising.
View increased talk time or increased chat handling time as indicators these escalated interactions are more complex and more emotional now. That means veteran agents and new hires need more training and coaching to handle those issues.
And here is the irony. The better your agentic AI is at containment, the better your human agent’s emotional intelligence and problem-solving skills will need to be.