HexClad's AI playbook: What it actually takes to make AI work in CX

At our recent AI Lab in LA, we sat down with Andrew, HexClad's Global Head of Customer Service, to talk about what it actually takes to make AI work in CX. Not the theory. The real, messy, behind-the-scenes work.

Andrew has spent 25+ years in CX and CS, leading teams at PayPal, Stamps.com, and Bed Bath & Beyond. He's done everything from answering phones as an agent to running orgs of 150+ people.

Here's what he shared.

He walked into a team that was curious but had no plan

When Andrew joined HexClad, Siena had already been in place for about eight or nine months. The automation rate was sitting around 10-12%.

The problem wasn't the tool. It was that no one had built a clear strategy around it. There was curiosity but no plan, no cross-functional buy-in, and no one treating AI as something that needed real investment.

So Andrew started from the ground up. Understanding the company's goals, identifying blockers, and figuring out which teams across the org were willing to partner on making AI work.

He stopped treating AI like a side project

One of Andrew's biggest points: AI is not something you plug in and walk away from.

"There is an idea out there that if you bring AI into your organization, you just bring it in and say, okay, I'm done. Go do your thing. That's not how it works. If anything, you're actually doing more work with AI because you have to put guardrails in place."

He described it as a full-time job. Weekly meetings with the Siena team, constant iteration on workflows, testing guardrails, and refining what the AI can and can't do. The payoff is real, but it's not instant.

The numbers tell the story

By Q4, Siena had become one of the top performers on Andrew's team by tickets closed. The impact:

• 45 fewer seasonal agents needed during peak season (October through January)

• 10,000 person-hours saved

• CSAT went from the 70-80% range to consistently mid-90s

Those numbers did something important beyond saving money. They got the rest of the company to take AI seriously.

"Being able to share that internally and say, here's how much we're saving and here's how much performance we're gaining. People started to buy into the concept of AI at that point."

How AI exposed HexClad's broken backend processes

This was one of the most honest takeaways from the conversation.

Andrew was blunt: AI will show you everything that's wrong with your operations. If you have a messy fulfillment process, AI won't fix it. It'll make the mess more visible.

"It forces conversations with other department leaders, and it forces awkward conversations at times to say, we're doing this all wrong. We have to clean this up."

His automation rate is at 35% right now. Not because of Siena's limitations, but because HexClad has a complex backend with multiple 3PLs, carriers, and tools like Parcel Lab and Fulfill that all need to work together. Until those internal processes are cleaned up, automation can only go so far.

Once they get the warranty workflow automated with Siena's new video recognition capabilities (which can distinguish between a chip, a scratch, overheating, and user error), Andrew expects to hit 60%.

He reorganized his entire team around AI

Before Siena, HexClad's agents were handling everything. Simple order status questions alongside complex warranty claims. There was no structure.

Andrew pushed tier-one tickets to Siena (where's my order, address changes, returns) and retrained his human agents on the work that actually requires judgment. Warranty claims, product education, nuanced customer conversations.

The result: he changed what he measured, too. Speed and containment for AI agents. Quality for human agents. Each measured on what they're good at.

"Agents were concerned at first. But taking away the busy work and giving them better things to work on, stuff that uses their brain and their judgment, it makes the job more interesting. The churn is less now than it was."

He even promoted Brianna, who had been the main point of contact with Siena from the beginning.

His advice for getting started

Andrew's number one recommendation: get buy-in from your stakeholders before you implement anything.

"If you're the one advocating for AI, you need to lead that within the company. Most companies don't have a dedicated AI officer. You by default become that person."

He also emphasized following the data over opinions. When he was deciding whether to loosen guardrails on social media or expand what Siena could handle, he didn't listen to the loudest voice in the room. He looked at engagement rates, CSAT scores, and customer feedback.

"The numbers don't lie. But people always have an opinion on this stuff."

What's next: Voice of Customer and a Shopping Agent

Andrew is now looking at Siena for voice of customer insights. Giving product, marketing, finance, and content teams real data on how customers are reacting to products, what's happening on social, and what patterns are emerging across channels.

He's also working with HexClad's growth marketing team on a shopping agent. A concierge-style experience on the website that can walk customers through product selection and checkout.

"We get a lot of tickets into CX asking those questions anyway. To have a proactive widget on the website that can walk people through that and take them through the checkout experience, that will save me a fair amount of resources."

The partnership matters more than the product

Andrew brought this up unprompted and came back to it multiple times.

"The reason I'm still with Siena is because of the people within Siena."

He's worked with a lot of vendors over the years and said most relationships are transactional. You're paying them to do a thing, and they do it. Siena was different.

"I get questions every week like, what is your North Star? What are you trying to achieve? What is the problem we're trying to solve here? Rather than just me saying, I need you to do X, Y and Z and do it as quickly as you can."

He described weekly meetings that regularly run long. Not because of problems, but because of ideas. What if we tried this? Have you thought about that? Those conversations push HexClad to think bigger about what's possible.

Andrew also pointed out that a lot of the Siena team comes from CX backgrounds, which makes it easier to speak the same language and solve problems together. When something goes wrong, a hallucination, a bad response, his team flags it in a shared Slack channel and it gets fixed almost immediately.

"That was very unexpected to me, and it actually helped build trust in them, which is why I've now used them at two companies."

He runs 115 agents across three BPOs, plus a small internal team of three people managing the AI program. The Siena team functions as an extension of that group. Not as a vendor on the other side of a ticket queue.

Why he chose AI-native over help desk AI

When asked why he chose Siena over his help desk's built-in AI, Andrew didn't hesitate.

"Help desk. Jack of all trades, master of none. Siena is all they do. You want experts in their field helping support your organization."

He also made the point that he doesn't want to be locked into a walled garden. He wants partners that do one thing really well and can plug into a bigger picture.

The AI Lab is our intimate roundtable series where CX leaders share what's actually working (and what's not) with AI. We host them in cities across the US. If you want to join the next one, stay tuned here.