What You Need To Know About Building a Conversational AI Team
I recently wrote about how conversational AI has the potential to change the way you do business and build relationships with customers. Now, I’m going to walk you through the steps for putting a conversational AI team together. I’ll cover who you need at the table and which skills are essential off the bat. I’ll also go over a few key ways that conversational AI teams might differ from your existing software teams. Finally, I’ll give you some tips for how to take an existing team to the next level with specialized talent.
Before I dive in, let me say a few words about using conversational AI in internal tools. Unless you’re a really large organization with very specialized functions, you’ll be best served by using existing solutions instead of building your own. For instance, instead of relying on email you might consider X.ai to automate meeting scheduling, or Skipflag to build an up-to-date knowledge base from your interactions in messaging apps.
While you don’t need to do custom development, I recommend you give internal conversational AI tools a try. Not only can they make your employees more productive, but they’ll also give your team a feel for what’s possible in a great conversational experience.
Software Development Teams with a Twist
When it comes to customer-facing conversational AI applications, you’re more likely to get value from investing in custom development. Chances are, you’re already shipping software with a team structure and process that works for you, involving product management, engineering, design and QA. The good news is that you’ll be able to use a similar structure and process for conversational AI purposes. However, there are some key tips to keep in mind:
1. Hire a conversation-friendly product manager. Using high-quality conversational AI platforms, you can make a lot of progress with a general purpose engineering team. But because conversational interfaces have unique challenges and opportunities, I recommend using a product manager with previous expertise. Look for someone who has built their own chatbot, is a heavy user of messaging and conversational AI products, and understands some basic tenets of natural language processing.
2. Design for conversation. Similarly, conversational design has several aspects that differ from other types of interface design. These include persona design, dialogue flow design, and conversational repair design. For speech, defining the right voice, pausing and earcons for the conversation are also relevant. If you can get a designer who already has some experience with conversation, that will help. Otherwise, you might need to build in time for your design team to ramp up on best practices in this area.
3. Personify your business (with input from the marketing team). Since users treat conversational agents like people, the personality of your chatbot or voice interface really matters. Even if the goals of the product aren’t specifically marketing-oriented, you’ll want marketing at the table to make sure your persona design aligns well with your brand.
4. Customize your QA process. Quality testing for conversational interfaces can be more complicated than for traditional web or mobile interfaces, for a couple of reasons. First, the range of responses a user can give (when natural language input is supported) is much less constrained since they can say whatever they want instead of choosing from a limited menu of inputs. Second, your conversational AI application is likely to rely on several machine learning models. You’ll need to make sure your testers can handle assessing the performance of your system quantitatively, or add a data scientist with complementary skills to help out.
To round out your conversational AI team, decide who is going to be responsible for supplying and adjusting the content you’ll be communicating in your conversational AI experience. This might be sales, customer service, or community management. Be sensitive to fears that AI may threaten jobs by thinking through how the experience will free up these teams for higher-level work or make their jobs easier. The emphasis should be on collaboration and empowerment, rather than competition.
Take Your Conversational AI Team to the Next Level
As conversational AI starts to become core to your business, consider supercharging your team with more specialized skill sets associated with particular roles. Let’s take a look at what these roles can offer and how to know when to bring them on board.
There are many different flavors of data scientists, but they’ll all be using your data to answer questions or solve problems. In general, once you have a significant amount of data flowing through your conversational AI system, you might want a data scientist (or several) to analyze and improve its performance. For instance, you might task a data scientist with:
- Segmenting your users into groups based on their behavior.
- Analyzing the other data you have about a user to inform future conversations.
- Understanding how people are using your conversational AI application.
Data engineers do the heavy lifting to manage your data storage and get it aggregated, cleaned up and ready to use in your conversational AI application. You might need a data engineer if you have data spread across a lot of different backends or if your data scientists find they’re spending more time cleaning data than actually analyzing it or using it in models.
Computational linguists / natural language processing scientists
Although terminology varies, I’m talking about someone who has a deep knowledge of linguistics combined with technical chops. Consider adding someone like this to your team if you’re building a broader conversational experience vs. one with only a few options.
You’d also bring people in this area on board if improving your performance on key language technology components would yield meaningful business results. For instance, let’s say you’re building a customer service experience that routes users differently if they’re unhappy vs. happy. In this case, you want to invest in someone who deeply understands sentiment analysis and can build a really accurate classifier.
When making this hire, be sure to focus on candidates with a background in the languages, channels and modes (text or speech) that are most important to your conversational AI strategy.
General machine learning experts
In addition to language technology, a successful conversational AI experience often has other machine learning components. For instance, a classifier might decide which channel to use when contacting someone or which product to recommend to a given user in chat. It’s worth adding expertise to the team if you’re going to initiate messages with users (instead of just responding to them) or if you’re building a heavily personalized experience.
Conversational UX designer / researchers
If your conversational AI experience isn’t meeting your goals for user engagement, you might invest in a designer with previous background in designing conversational experiences for text or speech. Designers with a background in call center or IVR applications or video game dialogue often have highly relevant skill sets.
For some conversational AI applications, you’ll have access to ample training data. For instance, you can develop content for a customer support chatbot from your existing customer support conversations. However, if you’re starting a totally new type of conversation, an experienced writer might be required to handle what you’re going to communicate in a way that is consistent with the persona of your bot.
Competition for talent in these areas can be fierce, so you may need to get creative. While you might find hires from academia, consider looking to post-graduate certificate programs or other alternative paths as well. Post a challenge to Kaggle, or attend conferences as a way to engage with the community. Make sure your recruiters are up on the latest language in the field. Eventually, you might even want to think about working with a dedicated AI-focused recruiter.
Are You Ready to Build Your Conversational AI Team?
Hopefully, you now have a good sense of who to call on to start integrating conversational AI into your business. You can dive into conversational AI for employee efficiency with a minimal team and off-the-shelf products. As you move into customer-facing applications, you don’t need to wait to get started until you have a specialized team in place. Start with people you already have, and add only as it becomes necessary.
In my next post, I’ll get into the nuts and bolts of how to start your conversational AI projects. In the meantime, I’d love to hear what your conversational AI team looks like and any lessons learned along the way. Connect with me on Twitter (I’m @jbrenier).