How Odido gave 1,800 customer operations agents their own AI coach with EasyCoach

Odido’s training team supports around 1,800 customer operations agents and a commercial workforce spread across mass market and retail. For Daan Stobbe, who has managed training and development at the company for nine years, the job has always been about one thing: making sure people are ready to handle whatever a customer throws at them.

For years, that meant doing in-person roleplays. Trainers would step into scenarios themselves: playing the difficult customer, the frustrated caller, the person who wouldn’t take no for an answer. It worked, but it had limits. Only a handful of agents could practice at a time. The feedback depended on who was in the room. And for the trainers, it was exhausting.

Easygenerator has been part of Odido’s learning setup for around six years. The L&D team has used it to build e-learning for the full organization, from compliance modules to product training, fulfilling 100% of training requests with just 25% of its original team. EasyCoach came later.

“The Easygenerator team showed us EasyCoach and asked us what we thought. We were totally surprised by an innovation like this. We didn’t know we had that need,” says Daan.

Roleplays have always been a cornerstone of Odido’s new hire program. Customer service agents handle complaints, objections, and frustration every day. Soft skills matter. So does knowing how to handle a cybersecurity threat from a caller trying to extract account information. The only way to build those skills is practice.

The problem was access. In a classroom, a trainer plays the other side of the conversation while the rest of the group watches. In the old setup, most agents got limited time on the floor before facing real calls. “There were only one or two people who could train the roleplay,” Daan explained. “Now everyone can—without having an audience.”

Feedback had its own complications. Trainers knew their agents. That familiarity is valuable in many ways, but it could affect the coaching. “Sometimes when you know the person, that can make you a little bit biased,” Daan said.

For the trainers themselves, running roleplays was draining. Staying in character, managing the room, and coaching in real time left little energy for effective observation. They were too busy being the customer to notice how the agent was actually developing.

The L&D team builds all the scenarios based on the conversations that come up most often in customer operations. That includes customers with objections and resistance, and cybersecurity situations. For team managers, there are separate scenarios: difficult performance conversations, job interviews, handling a team member who isn’t meeting targets.

Odido’s customer service agents practice using voice. They pick up a headset and talk. The AI plays the other side of the conversation, responds to what they say, and delivers feedback at the end of each session. On average, agents go through a scenario two to three times per session.

The shift for trainers has been just as significant. Instead of being inside the roleplay, they can now watch it from the outside. “They are a lot more relaxed when they can use EasyCoach and they can do their observations while learners are talking, instead of being part of the roleplay themselves,” Daan said. “When you do an in-person roleplay, it costs a lot of energy.”

Odido is in the early stages of measuring EasyCoach’s impact against its main operational KPIs. For now, the biggest shift is confidence, Daan said. Agents go into live calls having already run the scenario multiple times. They have heard the objections. They have practiced the response. They have received feedback that doesn’t carry the social weight of a manager’s opinion.

The objectivity of the AI feedback matters to agents in a way that peer feedback sometimes doesn’t. In a classroom, an agent knows the trainer. They might soften a critique or read the room. EasyCoach doesn’t.

“It’s a very objective feedback you receive afterwards. It’s very constructive and gives very good feedback to get better and better, to improve yourself,” Daan explained.

EasyCoach also changed what learning looks like at Odido more broadly. “The most important part is that we attract learners to learn,” Daan said. “Before they used to go through a very long e-learning, and now they can use EasyCoach and they are very attracted to it. They want to keep improving.

One application emerged that the L&D team hadn’t fully anticipated: using EasyCoach during hiring assessments.

When Odido holds selection days for new customer operations roles, candidates now complete EasyCoach scenarios as part of the process. Everyone goes into the same room with a headset and works through the same conversation. “You can put them all in one room, give them a headset, and just start talking,” Daan said.

A group of candidates can be assessed simultaneously, rather than one at a time with a trainer. Every candidate faces the same scenario, the same customer, the same criteria. The feedback isn’t shaped by which trainer happened to run the session.

This is now a standard part of how Odido evaluates candidates for customer-facing roles.

EasyCoach currently runs in customer operations and mass market retail. HR is exploring roleplays for operational management.

The balance between in-person and AI-led practice currently sits at roughly 50/50. Daan expects that to shift. “Our use of EasyCoach will increase compared to the in-person roleplay because the tool is also getting better and better,” he said. “Then we can use it a lot more than now.”