Siena's March AI Lab in NYC

What we talked about at Siena's March AI Lab in NYC

Every month, we bring together CX leaders from consumer brands for an honest conversation about what's working with AI, what's not, and what's next. Our March AI Lab in NYC brought together teams from Kitsch, Magic Spoon, Harry's, Oneskin, Caraway, Grüns, Deeps, and more.

Here's what came up.

Nobody asks "does AI work?" anymore

Three years ago, that was the question. Now the conversation has moved to: where else can we apply it? How do we improve what we already have? What are we missing?

One CX leader put it well: customers used to say "there's no way AI could do that." Now they say "I'm surprised it can't do that yet." Expectations have flipped. Customers expect more from AI, not less.

Start small or fail fast

Every CX leader in the room agreed on this: trying to automate everything at once is the fastest way to fail.

The ones seeing real results started with one or two high-volume, simple-resolution use cases. They refined those, built confidence internally, and then expanded. The ones that struggled tried to throw AI at every process, including complex ones like warranty claims that need photos or multi-step reviews. That's where things break down and bad public reviews start piling up.

The advice from one team that's been running AI for almost three years: start with the actions customers already want to do themselves. Order cancellations, subscription pauses, tracking updates. The interface is just conversational now instead of a portal. Customers get what they want in 10 seconds at 4am. That's a win.

AI will expose every gap in your documentation

Multiple teams shared the same experience: they thought their knowledge base was solid until they implemented AI. The gaps showed up fast. AI exposed every inconsistency, every missing edge case, every policy that wasn't documented clearly enough.

The good news: AI also helps you find those gaps faster. Once you see where the AI struggles, you know exactly what needs fixing. The teams that treated their AI setup as a documentation audit, not just a product launch, got to a strong foundation much faster.

The interesting part is what AI teaches you, not how many tickets it closes

This was the biggest theme of the day.

One team uses AI to surface product issues days after a launch. Instead of waiting weeks for patterns to emerge through manual review, they can ask: how many customers mentioned this issue? How frustrated are they? Is it trending up? They walk into product meetings with data, not anecdotes. And product teams actually listen.

Another team set up scheduled AI reports that run weekly and go straight to Slack channels for cross-functional partners. When they launched in a major retailer, their retail team started getting weekly voice of customer reports automatically. No one had to ask CX for anything.

One CX leader described how they use AI to validate whether something is a real trend or just a few loud voices. They can check the frequency, the intensity, and whether it's compounding with other issues. That kind of visibility used to require a dedicated analyst. Now it takes a question.

Giving everyone a license doesn't mean adoption

Multiple brands shared the same lesson: giving everyone an AI subscription and saying "go play" doesn't work.

One team tried it with ChatGPT. Adoption was inconsistent because there was no shared context, no company data loaded, no structure. The second time around, they connected their knowledge base to Claude, deployed it in Slack where it felt familiar, and seeded a small group of power users before opening it up. Those early users posted questions publicly, people saw the quality of responses, and adoption happened organically.

Another team took a similar approach: built internal tools on top of Claude, controlled permissions centrally, and let people request new capabilities as they discovered use cases. The key was making it easy and non-threatening. Nobody was forced to adopt. They just saw it working and wanted in.

Treat your AI agent like a new hire

Nobody in the room treats AI as set-it-and-forget-it. The teams with the best results run regular audits, test iterations, review edge cases, and have feedback loops where human agents flag anything the AI got wrong.

One team dedicates about 15 people across their CX org to a mix of auditing, improving, building, and testing. That's not 15 people just watching. It's 15 people operating the system, launching new capabilities like voice, and making sure quality stays high.

The consensus: AI is another team member. It needs coaching, feedback, and accountability, just like anyone else.

Do customers actually like it?

The room was honest about this one. Customers don't love AI when it adds friction, interrupts them, or can't solve their problem. The hotel AI that makes you say "agent" three times before connecting you to the front desk? Everyone hates that.

But when AI handles a simple action quickly, customers are happy. Fast resolution at any hour, no waiting, no guilt trip on a cancellation. One team shared that customers regularly leave positive reviews after interacting with their AI agent, calling it "really efficient."

The key principles from the room: be transparent that it's an AI. Make it easy to get to a human. Don't automate emotional or complex situations. And sell the value to the customer: "I'm here to get you the fastest resolution possible."

The role of CX is being rewritten

This was the thread running through every conversation. The role of a CX operator is fundamentally changing. Teams that used to spend their days on repetitive volume are now doing product intelligence, retention strategy, cross-functional reporting, and revenue-driving conversations.

One CX leader said it best: the shift isn't AI doing more. It's what your team becomes when AI handles the rest.

That's the conversation we keep having at every AI Lab. Not whether AI works. But what it makes possible.

Siena's AI Labs bring together CX leaders from consumer brands for real, unfiltered conversations about AI and customer experience. If you want to join a future one, sign up here.