A man rings his home insurer to start a claim. There’s been a fire. Halfway through reading out his policy number he loses the thread, apologises, says he hasn’t slept, asks the agent to repeat the question. Twice. The agent walks him through the form, books the loss adjuster, explains the temporary-accommodation cover, and closes the call inside the target handle time. Every step was done correctly. The man hangs up no better off than when he rang, because the thing he needed wasn’t a step.
He needed someone to notice he was in no state to take any of it in, to slow down, and to check he’d actually understood the one part that mattered. That isn’t on a checklist. It’s a skill, and it’s the hardest kind to teach, because you can’t write it down and hand it over.
So this post is about how you actually build that skill in an agent, rather than hoping they show up with it.
You teach it the way you teach any hard skill, with practice rather than instruction. Telling an agent to “show empathy,” or handing them a vulnerability checklist, doesn’t survive contact with a real, frightened customer. What does work is rehearsing the conversation itself, the bereavement call, the customer who can’t take it in, against a realistic scenario with feedback on the attempt, often enough that the first real one isn’t the one they learn on.
Why a rulebook doesn’t transfer
A vulnerability policy can tell an agent what counts as a signal and which team to escalate to. It can’t tell them how to sit in a silence without rushing to fill it, or how to ask “did that make sense?” so it lands as care rather than a test. Those are the things that decide how the call actually goes, and none of them fit in a process document.
Watch a genuinely good agent on a hard call and you can see the skills at work, even if they couldn’t name them for you. They slow down when they hear distress, instead of speeding up to protect their handle time. They acknowledge the thing the customer is frightened of before they start solving it, which is the kind of observable behaviour you can actually give feedback on rather than a vague instruction to be warmer. They check understanding without making the customer feel slow. They know the moment to put the script down, and they stay steady when someone is upset instead of going cold or getting flustered. Every one of those is learnable. Not one of them is teachable by being told.
The trouble with how we usually teach it
The best way to teach this has always been role-play with a skilled partner, someone who can play the distressed customer and then tell you where you lost them. It works. The trouble is it doesn’t scale. You can’t give every agent enough sessions with a good human partner, so most get a handful in their first week and then almost none.
Which means the real teacher becomes the job, and agents end up learning to handle vulnerable customers by practising on vulnerable customers. A new starter’s first bereavement call is a real bereaved person, and when it goes badly it goes badly for someone already having the worst week of their life. The feedback, if it arrives at all, turns up a fortnight later as “try to show a bit more empathy,” which is about as useful as telling someone to be funnier.
These calls are also rare for any one agent. A person might take a single genuinely hard vulnerability call a week. At that rate it takes years to build up the reps that make someone calm and good at them, and plenty of agents move on before they get there. The skill that matters most is the one your training gives the least practice at.
What practising the real call looks like
The way out is to give agents the reps without the cost. Voice role-play lets an agent practise the call they find hardest, out loud, against a customer who reacts the way a frightened or confused one actually does, and get feedback on the attempt straight away instead of three weeks later. They can run the bereavement call a dozen times in an afternoon, get it wrong where getting it wrong is safe, and try a gentler opening, before a real grieving customer is ever on the line.
A few things make this more than a gimmick. The difficulty can climb as the agent improves, so a confident agent gets a customer who’s genuinely hard to reach rather than the same easy script again. The agent can take the customer’s side for a round and hear how their own briskness sounds from the other end. And because it’s voice, they get to practise the parts that only exist in speech: a held pause, the pace they set, the tone that tells a distressed person they’re being listened to. No written module rehearses any of that.
The obvious worry about rehearsing a hard conversation is that it produces a rehearsed-sounding agent, someone reaching for a memorised empathy line that lands as insincere. That’s a real risk, and it’s a design problem rather than a reason to avoid practice. You aren’t drilling a script; you’re building judgement, so the scenarios have to vary and the customer has to be unpredictable enough that there’s no single right line to learn. Done well, the agent stops reaching for lines at all. What sticks is the instinct to slow down and check, and it surfaces differently on every call because every call is different.
Coach the specific gap, not the general weakness
The strongest version of this doesn’t drill agents on a generic “difficult customer.” It builds the practice out of the calls they actually fumbled. This is the coaching step of the closed loop we’ve written about: rather than a vague note that someone’s empathy scores have slipped, you take the exact call where the customer said “I’ve just lost my husband” and the agent carried straight on to direct debits, and you turn that moment into a scenario they rehearse until they handle it differently.
That precision is what makes the coaching stick, and it’s also what lets you know whether it stuck. Coaching that everyone enjoys in the room but that changes nothing on the next live call is just expensive reassurance. The only way to tell the two apart is to look at the agent’s later real calls and see whether the behaviour actually moved. The practice and the proof are two ends of the same thing, which is the whole point of running it as a loop rather than a training day.
It’s kinder to the agent, too
There’s a second reason to take this seriously, and it has nothing to do with the customer. Handling a run of distressing calls with no preparation and no feedback is one of the quicker routes to burnout on a contact centre floor. An agent who has practised the hard call and knows they can get through it is steadier on the live one, and steadier afterwards. That kind of steadiness comes from reps, not from the hiring panel.
This only matters more as routine contacts move to automation. Once the simple, scriptable calls are handled by software, the ones left for your people are the hard, human ones: the bereavement, the complaint, the frightened first-time claimant. The skill this post is about stops being a nice-to-have at the edges of the job and turns into most of the job. Which is a fairly strong argument for getting serious about teaching it now, while you still have a stream of routine calls to build the easy version on.
You can’t write it down, but you can practise it
The thing the man on the fire claim needed could never have gone in a process document. No amount of “remember to show empathy” would have changed his call, because the gap wasn’t knowledge, it was skill, and skill comes from doing the thing, getting it wrong, and doing it again. You can’t hand an agent that. What you can do is let them practise it fifty times before it counts for real. That’s the difference between hoping your people are good at the calls that matter most and knowing they are.
Want to see what this looks like on your own hardest calls? We’ll take a real, fully anonymised difficult call from your operation and build it into a role-play your agents can practise against, so the next frightened customer reaches someone who has already had the conversation.