How Tediber Achieved 64% AI Automation and Cut Customer Service Response Time from 72 Hours to Under 1 Hour with Yuma AI

Introduction

Tediber is France's first bed-in-a-box brand. Founded in late 2015 by entrepreneur Julien Sylvain and a team of co-founders that includes the design duo Juan Pablo Naranjo and Jean-Christophe Orthlieb, the company was built on a simple observation: quality mattresses made in France that were sold at a fair price online - was nearly impossible to find.

Julien, who is still CEO, saw an opportunity in the gap between overpriced retail and underperforming online options. His strategy combined strong online presence with a direct-to-consumer model, and it worked. Tediber has since sold over 270,000 mattresses and been significantly driven by positive word-of-mouth, collecting more than 80,000 customer reviews. The product line has expanded well beyond mattresses to include bed frames, sofa beds, pillows, duvets, bed linens, sofas, and a full children's range. The company operates 4-5 of its own retail stores across France and ships to most EU countries.

As the product line and customer base grew, so did the demands on Tediber's ecommerce customer support operation. Océane Turpin joined Tediber as Customer Care Director in late 2025, bringing over 11 years of customer care experience. She arrived to a team of 6, managing omnichannel customer support across email and phone, supplemented by OnePilot, a third-party customer support agency. Within her first weeks, she was tasked with evaluating and implementing Yuma AI during the Black Friday and Christmas rush.

"When I was setting up Yuma, I compared it with what I had to do to set up another AI tool at my previous company. Two completely different levels. Yuma is so much easier. If you have basic knowledge about technology, you can deep dive into the back office and start creating your own automations and flows." - Océane Turpin, Customer Care Director at Tediber

‍

The Problem

About a year and a half before Océane joined, Tediber hit a peak in sales. That growth was obviously good news for the business, but it created immediate pressure on the ecommerce customer support team, which at the time consisted of just four agents and no dedicated manager.

By then Tediber's product catalog spanned everything from full-sized mattresses to pillows and bed linens, and the company worked with five different shipping carriers across France and the EU: Edeliv for Paris, VIR for the rest of France, France Express, DPD, and DHL for international orders. Each carrier has its own tracking systems and delivery workflows. When order volume surged, shipping queries surged with it, and the team's existing IT tools made it difficult to track shipments across all five providers in one place.

On a typical Monday, the team could face 300 incoming requests. Many of those were about shipping status, delivery dates, or shipping issues - common ecommerce post purchase experience friction points. Others were straightforward invoice requests that piled up because Tediber's system wasn't sending invoices to customers automatically. On some days, that single query alone generated 100+ tickets.

"If you have, as a customer care agent, to create invoices yourself for 100 requests in one day, it's so long, and for nothing." - Océane Turpin, Customer Care Director at Tediber

The volume problem compounded. Because the team couldn't respond quickly, customers would follow up multiple times through email and phone, creating duplicate tickets for the same issues. Customer service response time stretched to 72 hours, and each unanswered message added to the backlog.

Meanwhile, incoming phone calls were increasing as well, and those took significantly longer time to handle compared to emails. The team needed breathing room on the email side so they could dedicate more time to phone support, where a human conversation mattered most.

To manage the overflow, Tediber brought in OnePilot, a customer support agency staffing human agents to handle email queries. But the arrangement came with its own costs. Even then, because Tediber's internal processes changed frequently, the outsourced team struggled to keep pace with the latest workflows and maintain the same quality as the core team.

The operations manager at the time, Elise Vettel, began searching for AI tools for ecommerce that could absorb the growing volume, give the internal team space to focus on complex and sensitive queries, and reduce the growing dependency on outsourced support.

‍

Solution

Tediber's first got to know about Yuma through a board member who had been evaluating AI solutions for another company. The person sent Yuma to Elise Vettel, Tediber’s operations manager, who booked a demo with Yuma. Before that, Elise had been working with Tolk AI, an AI chatbot for ecommerce she was dissatisfied with for quality reasons.

Tediber kicked off Yuma in late October 2025, during a trial period that happened to coincide with Black Friday and Christmas, the busiest stretch of the year. Océane, who had joined the company around the same time, was responsible for setting up and evaluating the tool during the trial, then reporting back on whether to continue with Yuma or not.

The timing was difficult. Océane was simultaneously learning Tediber's internal processes, managing her new team through peak season, and configuring AI at a company she had just joined. What made the difference, she says, was how the setup process worked compared to what she had experienced with other tools.

At her last company, Océane had implemented Volubile, a voice AI tool. That process required filling out Excel files with multiple tabs: one for intents, another for API integrations that needed IT security review, and one for writing out every process command in a single unsorted list. There was no way to organize the process library by topic, no visual interface, and no guided support. All in all, the set up was very complex and hard to navigate. Comparing that to Yuma, this is what she had to say:

"When I was setting up Yuma on the back office, I realized it is so much easier. You have so many settings, but they are so easy to access and update. Even the Playground and the Ask Why features, where you can deep dive into a ticket and see how it was handled by the AI, that is not something I saw in other tools." - Océane Turpin, Customer Care Director at Tediber

Nour, Yuma's Account Manager for Tediber, handled most of the initial configuration at launch and remained closely involved throughout the trial. Océane describes her as "a master of the Yuma back office" who responds within the same day, communicates her availability clearly, and builds solutions while teaching Océane how they work so she can learn to manage them herself too.

"The onboarding during the trial was very well organized. You could definitely sense the quality of work, and the potential that the AI could have in a few months in terms of automation." - Océane Turpin, Customer Care Director at Tediber

Tediber's Yuma AI agent (the team named it Sébastien) connects directly to Shopify and operates within Gorgias, their ecommerce help desk. Océane manages the back office, creating and updating autopilots, processes, and customer knowledge base. The rest of the team interacts with Sébastien when it escalates tickets to them, either because the query requires human judgment or because the customer has expressed emotional dissatisfaction.

When a conversation requires a concession, a nuanced reading of the customer's frustration, or a decision that falls outside the configured processes, the ticket moves to a human agent. Océane reviews escalated tickets regularly to identify where Yuma’s knowledge or workflows need updating, which she treats as a continued responsibility.

"AI is great, but it's not going to replace our job. If we improve the AI agent to the point where it can solve most of the basic stuff, then it's just going to create more time for the humans, and the humans can invest this time to increase the quality even more." - Océane Turpin, Customer Care Director at Tediber

Four months in, Océane says she is now hitting her stride with the tool, because she finally understands Tediber's processes well enough to know exactly what she wants Yuma to learn next. Her recent focus has been on the Flow Builder, which she sees as particularly useful for configuring more complex workflows.

‍

Outcomes

64% of Customer Queries Automated

Four months after implementation, Yuma is handling 64% of all incoming customer queries through automated customer service. The most common automated queries include return requests, product issues, shipping status and date inquiries, exchange requests, pre-sales and product questions, and shipping changes. Many of these, like the invoice requests that once consumed entire days of agent time, are now resolved without any human involvement.

72 Hours to Under 1 Hour in Customer Service Response Time

Before Yuma, customers had to wait up to three days (72 hours) for an initial reply. That delay triggered a cycle of follow-up emails and phone calls, which created duplicate tickets and added even more pressure to the backlog. With Yuma responding to a significant share of incoming queries, the first response for those tickets now arrives within an hour (instant for simple replies and up to 1 hour if human is required). For the team, fewer duplicates and a lighter queue means they can also respond to their assigned tickets faster.

"A few months ago, the first answer from a human would take 72 hours. Now, for a large portion of tickets, the answer is within an hour. It's a real undeniable difference." - Océane Turpin, Customer Care Director at Tediber

~€900 in Weekly Cost Savings Against Outsourced Support

One of the clearest financial indicators came from comparing Yuma's output against OnePilot, Tediber's third-party support provider. OnePilot charges per email ticket and when compared - In a recent week, they saved roughly €900 in outsourced email costs for that week alone. As Yuma takes on more query types, those savings continue to grow, and Océane's plan is for Yuma to progressively take over the volume that would otherwise go to OnePilot.

Reduced Workload and Improved Team Morale

Before Yuma, Monday mornings could mean 300 incoming tickets for the team. After implementation, that number dropped by 33%. Océane describes the psychological weight of a large ticket queue as a real factor in her team’s wellbeing: when team know the volume is manageable, they have more time and focus to invest in ecommerce customer service quality and quality of customer relationships. The team also knows that Yuma is handling the repetitive queries, which frees them for the work that benefits most from a human touch, particularly phone support and complex product issues.

"Yuma is helping to decrease this workload, which helps on the mentality of the team as well. They know they're going to have more time to focus on their own tickets, because they know Yuma is taking the other ones." - Océane Turpin, Customer Care Director at Tediber

Responsive, Hands-On Customer Success

Océane credits much of the smooth implementation to Nour, her account manager at Yuma. Nour is known to respond within the same day, communicate proactively, and build configurations while explaining how they work so Océane can manage them independently. For Océane, who has worked with account managers at other companies, this level of engagement stands out.

"She's a master of the Yuma back office. I know that if I have issues with something, she's going to create a solution for me and help me understand how she did it, so she can share her knowledge. This has been an amazing experience." - Océane Turpin, Customer Care Director at Tediber

‍

Tediber’s Advice for Companies Considering AI for Ecommerce Customer Support

We asked Océane what advice she would give other companies evaluating AI for ecommerce customer support. Two recommendations stood out:

Get your internal tools in order

Before asking an AI agent to pull order data, check shipping status, or generate invoices, the systems behind those actions need to be in order. Océane points to Tediber's own order preparation tool as an example: it is an older system without API access, which means Yuma cannot retrieve order status from it directly. Because that tool also fails to pass correct information to Shopify, order tracking is one area that needed fixing. Her advice is to audit and fix your core IT stack before implementation, because the AI is only as capable as the systems it connects to.

"My first advice is to have your tools, your Shopify, your emailing tools like Klaviyo, all the rest of the IT working well. If you have issues inside your IT stack, the AI is not going to be able to find an answer for the customer." - Océane Turpin, Customer Care Director at Tediber

Dedicate someone to own the AI Implementation

Océane's second recommendation is to assign at least one person on the team to manage the AI automated customer service implementation process and also to manage the AI going forward. That person needs to review escalated tickets, understand why they were escalated, and then create updates, corrections, or new flows in the back office. Without someone doing this consistently, the AI won't improve over time.

‍

Conclusion

Tediber's customer support operation is still early in its AI journey, four months in, and Océane is the first to say there is more to build. She's expanding Yuma’s coverage into product quality workflows using the Flow Builder, gradually shifting email volume away from OnePilot, and refining the customer knowledge base as she deepens her own understanding of Tediber's processes.

What makes Tediber's story worth paying attention to is the conditions under which these results were achieved; and what it reveals about the practical potential of generative AI in retail and ecommerce. A new CX leader, in her first month at the company, implemented an AI agent during Black Friday and Christmas, with a lean team and complex multi-carrier shipping operations, and reached 64% automation within just four months. Customer service response time went from three days to under an hour. Weekly cost savings against the outsourced provider are already measurable.

For brands facing similar pressure (growing order volume, strained ecommerce support teams, rising outsourcing costs) Tediber's experience suggests that the right AI solutions for retail and ecommerce can deliver meaningful results quickly, even under less-than-ideal circumstances.

‍