Your CxD center of excellence (and why you don’t need one right now)

The conversational AI industry is notoriously difficult to define. Like a cresting wave, it seems like every time we discover a new “best practice” or a simplified way of working, it’s washed away by the influx of the new. So when the expectations from customers are high and industry-wide best practices are still taking shape, it’s hard to imagine creating something like a conversational AI center of excellence (CoE). Because, what is excellent right now?

But let’s define this term. “Center of excellence” is a pithy buzzword that’s come to mean different things across industries. For tech companies, a CoE is often defined as a team that provides best practices, governance, and support to the rest of the organization. They help their fellow teammates be great at that thing, usually in the form of performing training, providing support, standardizing processes, and solidifying governance. For example, a writing center of excellence.

If you’re leading your own CxD team, it’ll be some time until our industry can mature enough to build true centers of CxD excellence. Which is why I’m here to let you off the hook. In the meantime, the goal for CxD leaders should be to develop intentional best practices and governance right now, while creating forward-thinking process documentation. This article is about how to begin thinking in this way—one foot in the present and one foot in the future of your budding CoE. 

The four things you need to begin building your CxD center of excellence 

Developing a CoE within your team means you are equally focused on how to achieve success in your current projects, and how you might teach those exact skills to someone else, or replicate successful results later on. You do this by solidifying the good processes you already have in place and refining the others that are in flux. 

So, if you’re building your CxD team, expanding your team’s output, or you just want to become the kind of team that creates repeatable success, here are four strategies that’ll help. 

1. Learn from anyone who will talk to you 

When I started as a CxD leader at Instacart, I knew I couldn’t apply everything I’d learned at GE or Rocket Mortgage to the new team I was building. Every single CxD team speaks a different language and has different goals and ways of achieving them. For instance, I found that well-documented governance was one of the most important things I could do in my new role. And the way I discovered that was to talk to anyone who could spare 15 minutes. 

I began by looking up people's titles—UX researcher, design lead, operations, even people managers—and scheduled meetings with each of them to gather information. I wanted to know how they built their governance, processes, and operational structures. I asked how they defined “center of excellence.” I even asked how they decided which practices were best—how they enshrined them and made them repeatable. 

I read internal research, engineering and process documents, scoured our Confluence libraries, Wikis, and Notions. Then I met with the people who developed those resources and asked them my questions, so I could recreate them for my CxD team. I also asked who I should meet next—who is doing a phenomenal job of building their team and their own center of excellence?

I had two conversations a week for six months with folks across the organization. I got hand cramps from all the notetaking. But the results were astonishing. One conversation with the leader of the people operations team led to me borrowing their spreadsheet templates and reworking them for our own use. They were so organized and their processes have helped inform our own—something that wouldn’t have happened had I not sat down to learn from a more established team. 

This approach allowed me to also get to know the folks I would be collaborating with—especially engineers and product managers. I knew that if I learned from them (and even made a few friends), our projects would go smoother and make our lives a lot easier. I turned out to be right. 

2. Don’t skip defining clear roles and expectations  

I’ve had plenty of experience with conversation design roles that weren’t defined or didn’t have expectations.  Honestly, that kind of role kinda works for me. But for most people, a vaguely defined role paired with unclear expectations could mean that an entire team of CxDs feels untethered. I've been in scenarios where two or three conversation designers are working on competing priorities and no one can decide whether there’s a singular path forward. A shifting conversational AI industry shouldn’t mean there’s a constant state of internal confusion in your CxD team. 

Building a one-pager with the roles and responsibilities of each person on the team frees everyone to focus on what they do best. Even if every day feels like a different job—and as CxDs, we know it often does— there should be basic expectations for each role. Only from there can we add tasks to each role that contribute to building a center of excellence—like creating process documentation or a style guide (more on that in part 4). 

3. Choose the right tools 

When I started out in CxD, I used whatever tool I thought would be cool that week. It'd be like buying a new notebook. For a time, half my projects were in Miro and the other half were divided between Voiceflow and Excel. As fun as it was to try a bunch of new tools all the time, it’s hard to scale a cohesive CxD team when you don’t have the right tools.

A big part of the center of excellence is finding your tools and creating a shared library of resources to help your designers fully utilize those tools. In the end, I tried 15 to 20 different tools before landing on the ones we’d use regularly. And I still dedicate a couple of hours a month to trying out new tools because our space is changing so much. 

My advice? There are a few industry leaders. But not every tool will work for every team and use case. Find yours and build training, support, and process documentation around those tools. Your goal is to create an operation that enables a new designer to join your team and asynchronously and independently get up to speed on your tech stack as soon as possible. 

4. Create a living style guide 

A style guide is essential if you want to maintain brand consistency. If your AI assistant reads as if it’s been written by several different people, you’re doing something very wrong. The most obvious way to avoid this is to standardize certain stylistic rules—dates, numbers, contractions, and word choice. That portion of the style guide should be a ready-made checklist anyone can follow. 

But the most important part of the style guide is the voice and tone section. We even have a persona for our bot. So for a hypothetical example, this bot is a 34-year-old, gender-neutral person who grew up in the Midwest, whose major characteristics are care and empathy. Their humor sounds like an elder millennial, but they never make pop culture references. Maybe they're helpful, but they're not a pushover. They're confident, but not arrogant. This level of detail and direction on voice and tone will keep your team on track. We created the style guide as a main document that supports our CoE—it's written and designed to help us create one cohesive persona even though we’re all working on disparate parts of the AI assistant.

Style guides take a lot of people to create—hence why building internal relationships is paramount. Our UX research team defined our target audience, our marketing team helped us edit the guide for clarity, and customer success helped us refine it with user feedback. We relied on the real-life experiences of users and internal team members to understand how our assistant should speak and react. We’re constantly tweaking the style guide as our best practices evolve. 

Spread the excellence around

In our continued attempts to build systems, refine processes, and document our successes, we can’t forget that our way isn’t the only right one. Whether it’s gut-checking our ongoing projects using peer reviews or swapping stories with other conversational AI teams at networking events, we all know that in the pursuit of excellence, we all have something to offer and to gain. To give and to take. 

Conversation design is an ever-evolving beast—the rise of generative AI, the increasingly sophisticated expectations of customers, and access to technology means we’re entering a new era. That’s why starting to create the processes and documents that lead to excellence is something you should be doing right now. 


What Nike's chatbot taught us about conversation design best practices

No items found.