Endy is one of Canada's leading sleep brands, and part of a family of ecommerce companies that includes Casper and Hush. Together, these brands serve customers across North America, each with its own identity, its own voice, and its own expectations for what a great customer experience looks like.

When Erin Gray, Senior Vice President, Customer Experience, chose to invest early in AI, it wasn't a reaction to industry pressure, it was a strategic bet. "Our goal was to get ahead of something we knew would become table stakes for any CX team," she said, "and to grow our operations alongside the brand."

Endy had already made the leap from scripted chatbot to AI agent with Ada, but Erin knew that was just the beginning. She saw what AI could unlock with the right focus and built the team to match the ambition, bringing on Huda Shaalan as Team Lead, CX AI & Automation, with a mandate to take the AI customer service program further.

What followed was a year of deliberate, disciplined agentic transformation. Today, Endy's AI agent automatically resolves customer interactions across messaging and email for three brands—driving lower costs, stronger CSAT, and impact that reaches well beyond customer service.

This is the story of how they got there. Not only by deploying technology, but by building the operating model, practices, and partnerships that make AI a durable advantage.

Scaling with intention

Scaling AI isn't just about adding channels. It's about making your AI agent the first point of contact for every conversation, across every brand, every market, and every surge in volume. And that’s exactly what the team at Endy did.

From single-channel support to multi-brand AI operations

When Huda joined, the AI agent was already delivering on messaging channels for Endy. From there, the team moved with intent: expanding to email, deploying AI agents for Casper and Hush, and building out increasingly complex workflows such as cancellations, address changes, and routing high-risk tickets.

That pace for scaling was deliberate. Huda used automated resolution rate as her measure of readiness, a signal that the AI agent was performing well enough to be trusted on a new brand or channel. The AI agent needed to reliably deliver at least a 50% automated resolution rate on messaging before extending to new brands, and at least 40% on email.

"Within my first year, we expanded to email and expanded our processes—automating cancellations and address changes, scaling across two new brands and replicating the same successful workflows," Huda said. "Our automated resolution rate has increased by 20 percentage points since I first onboarded."

What made expansion across brands operationally smooth once those thresholds were met was how Endy had structured its Ada infrastructure from the start. Playbooks were standardized across all three brands, with differentiation layered on top: tone, greetings, sign-offs.

From day one, Erin insisted on a single, multi-brand relationship with Ada—one contract, one shared infrastructure—rather than the siloed, brand-by-brand approach common with other vendors. "We're a multi-brand organization, and we want one relationship, not one contract per brand. Some vendors would have pushed back, but Ada wanted to grow with us," she said. “Each subsequent expansion felt like configuration, not construction."

With three brands live across messaging and email, the team is now turning its attention to voice, approached with the same “start simple” mindset that's guided every launch before it.

The Black Friday/Cyber Monday test

Peak season is when AI customer service comes in clutch. Year over year, the Endy brand received 10,000 fewer tickets to live agents during November and December, freeing the team to focus on the complex, high-stakes issues that genuinely need a human touch. “When I pulled peak season results, I was pretty shocked," Huda said. “I honestly can't imagine what our team would have been like without Ada.”

And the number only tells part of the story. Those weren't just tickets closed, they were customers who received full resolutions, without follow-up contacts or escalations on the back end. With Huda's team actively managing AI agent performance, accuracy and safety are built into every interaction. "To know that the agent isn't just handling a large volume of tickets, but handling them accurately, correctly, and safely is what makes the difference," Erin said.

Address changes were a particular standout. Every Black Friday, a predictable surge of customers realizes their order is heading to the wrong address. The AI agent now handles those end-to-end, verifying identity, confirming the modification window, and executing the update through the Shopify API. No handoff required.

New market, minimal rework

The team was already mid-implementation on two other channel expansions. Ada's foundation made it possible to support a new market without missing a step.

“We were already mid-launch on Hush and Casper when we got the news about expanding shipping to the US, so we simply added it to the roadmap. Ada made it easy to customize for the new market without having to rethink our entire approach."

Valuable operational maturity

Playbooks that do real work

The address change workflow is one example of how Endy uses Playbooks: structured, multi-step processes that the AI agent executes autonomously from first message to final resolution.

When a customer reaches out, the AI agent verifies identity by matching their name and email against the Shopify order, confirms the order is within the modification window via a time-based tag, collects the updated information, and executes the change through the Shopify API.

Cancellations follow the same structure: verify, check eligibility, collect, execute. What makes these Playbooks valuable beyond their individual impact is that they're built to replicate; once proven on Endy, a workflow extends to Casper and Hush with minimal adaptation.

Data that changes the business

One of the less obvious returns from Endy's AI customer service program has been organizational leverage.

Address change is a high-volume contact type, and for a long time, it came with a hard constraint. The warehouse allowed only a limited time window after an order was placed to modify a delivery address. Once that window closed, the customer was locked in.

Customers who missed the window faced a frustrating dead end: the order would be sent to the incorrect address, and the CX team could only apologize. In the best case, a neighbor could hold the package. In the worst, it would be returned to sender and reshipped—a poor experience for the customer and a costly one for the business.

The CX team had been making this argument for years, but the warehouse wasn't convinced. Anecdotal accounts from human agents weren't enough to move the needle. After all, it was easy to counter that agents simply hadn't responded fast enough, even when customers reached out within the allotted window.

What changed was data. With the AI agent handling address change requests at scale, the volume of customers reaching out outside the window became documented, timestamped, and impossible to dismiss. Armed with that evidence, the team went back to the warehouse and won an extended modification window.

"We had data to show how many were coming in," Erin said. "The AI agent processes address change tickets instantly, which demonstrates that the customer needs more time to discover the error. We were able to double the length of that window."

The result was a better customer experience and less operational waste, as well as a clear illustration of what AI data can do beyond customer service. When CX has the numbers, it has the leverage to advocate for customers and change policy.

AI as a source of business intelligence

The address change win isn't the only example of AI data feeding back into the business. For Endy, the contact center has become a source of customer intelligence that no other team can replicate, and marketing is starting to notice.

When Endy secured a placement in Tim Hortons® Roll Up To Win™ contest—one of Canada's most iconic annual promotions, run by the country's largest coffee chain—their redemption code went out on 200,000 cups nationwide. It was a major brand moment, and Huda's team treated it like one. Before the campaign went live, they coached the AI agent specifically for the incoming surge: anticipating the questions customers would ask, making sure the agent could handle every related inquiry cleanly and accurately. When the contacts came in, the AI agent was ready, and it showed.

With the campaign wrapped and the data in hand, Erin brought it directly to the marketing team. "Marketing had traffic, sales, redemption rates," she said. "What they didn't have was visibility into how customers wanted to engage once they picked up that cup. If they're calling us, why? It was a level of engagement the marketing team wasn't expecting. They thought traffic and sales were the number one thing to track, and we were able to give them another layer of intel on top of that."

It's a pattern that shows up across campaigns and initiatives: CX data surfaces customer intent, behavior, and unmet needs that conversion metrics alone can't capture. For Endy, the AI agent isn't just resolving contacts, it's generating intelligence that makes the whole business smarter.

Making AI a company-wide effort

Endy brought far more of the organization into the AI program than most teams would think to. With three distinct brands, Huda tasked the creative team with developing an avatar, a name, and a persona for each brand’s AI agent. The company voted on names in a town hall.

"I encourage every brand to make it a team exercise," Huda said. "Sometimes the CX tool stack feels divided from what everyone else is doing. Looping people in provides visibility and makes it something the whole company feels ownership over."

The benefits of that visibility run in both directions. When teams across the organization understand what the AI program can do and feel invested in it, they start looping Huda in proactively. When the warehouse plans a sale, Huda's team gets the heads up. The AI agent gets equipped with updated product knowledge and new articles before the surge hits.

That kind of cross-functional evangelism is one of the clearest markers of a mature AI customer service program. Visibility doesn't just build goodwill, it makes the AI agent smarter, the team more prepared, and the business better positioned to capitalize on every customer moment.

Every drop-off is an input

Scale was the first chapter in Endy’s AI journey. Quality is the current one.

With multiple channels live across three brands, Huda's focus for 2026 is accuracy, consistency, and reducing the friction that prevents conversations from resolving cleanly. Testing standards remain high. Before any workflow goes live, every interaction must meet one clear bar: the AI agent's response needs to match what a skilled human agent would provide.

Ada's MCP integration gives the team the analytical depth to understand where drop-offs are happening and where workflows fall short. "We're leveraging MCP to really understand where those friction and drop-off points are,” Huda explained. “We have big CSAT goals for the end of 2026, and this level of visibility is necessary to achieve them."

Underpinning all of it is Erin’s mission to treat AI performance as an ongoing discipline that requires a committed team. Endy hired two QA analysts dedicated to AI agent performance, reviewing conversations across all three brands for accuracy, helpfulness, and on-brand tone. Their job is continuous tuning: catching errors, improving responses, and coaching the AI agent to keep it aligned with Endy's standards. Erin's pitch to leadership when she introduced the roles: "The best performers still need coaches. Even Rory McIlroy needs one."

Looking ahead, a warranty Playbook is in development and expected to handle a significant share of contact volume once live. Voice is on the horizon, approached with the same start-simple philosophy that's guided every previous expansion.

A team that grows alongside the program

Endy was clear from the beginning about what the AI customer service program was, and what it wasn't. "This was never a play to bring the AI agent in so we can reduce headcount," Erin said. "We pitched this to our human agents so they would be welcoming: this is there to help you do your job better."

The goal was always cost avoidance and capacity for growth, building a customer service center that could scale alongside the business. When volume relief arrived, the response wasn't fewer people. It was new roles: AI coaches, QA analysts, career paths that didn't exist before.

"When we've had some reduction in volume going to live agents, we say: now we have an opportunity to create a new path within the organization," Erin said. "That kept us true to our word. When the team sees us leveraging Ada for more, there's not the same suspicion. It really is creating a new center of opportunity."

Huda involves the CX team directly in decisions about what to automate next, having open conversations about where AI can help most and where careers go from there. The result is a team that has grown into the program, not away from it.

Best-in-class ecommerce CX

Endy's results didn't just come from deploying the right technology. They came from building the right program around it and maturing that program constantly.

The decisions compounded over time: deliberate thresholds, shared infrastructure, active management, cross-functional buy-in, and a partner built to grow alongside them. Each one reinforced the next, turning a strong foundation into a durable operating advantage.

For ecommerce brands thinking about what AI customer service can actually deliver at scale, Endy's story is the answer, not just to what's possible, but to how you get there.