Thoughts

CXD in 2023: reflections and projections

Braden Ream
|
December 16, 2022

We're entering the next phase of conversational AI, and it's incredibly exciting.

Conversation design as a practice is maturing, and the introduction of large language models (LLMs) is pushing forward the underlying technology at a rapid pace. The LLM "ChatGPT" by OpenAI had over a million users in less than a week. This renewed excitement for the conversational AI space paired with the rapid growth of conversation design teams is the start of a new era for conversational AI—one that delivers not only on cost, but amazing experiences.

These are exciting times. Here are my reflections on the year and projections for the next.

Large language models have entered the chat

I don’t believe GPT3 isn’t taking anyone’s job anytime soon because you and it do different things. It writes words. But your job was never about writing words—it’s always been about creating experiences. In an environment where there’s nobody to write for you, you probably tend to do a lot of writing. But if a machine can assist? That frees you to craft more and better experiences. 

I think large language models like this have the potential to make CXD roles a lot more data-driven. They’ll make the job more about curation, not creation. About ensuring the system meets user expectations more than crafting the things they interact with. It’s quite analogous to flying—pilots no longer land planes. But we have pilots. Their job is to get people from point A to point B. CXD will be similar. 

Overall I think it’s an exciting trend. Though, I think there’s a lot about the human experience it’s going to miss and it will forever need a human editor. 

Plus, when something suddenly becomes cheaper, a lot more of it happens. Just like how when banks launched ATMs, they actually hired more human tellers (though not at each branch), I think this is going to create a lot more work for all. 

We’re already looking for ways to implement large language models in Voiceflow—more on that next year.

CXD teams are maturing and hungry for standards

“What’s the best practice?” I’ll bet you hear that a lot. We think it a lot. Many teams have reached a size and maturity where they’re looking to organize around well-defined standards. They’re creating template libraries, documenting brand voices, crafting persona guides, and trying to figure out how to govern assistants that span many languages, geographies, teams, and cases. 

At a certain size, these teams are learning that the process matters more than the output. And that individual “superhero” team members who have all the knowledge in their heads and frequently save others might actually be contributing to the company’s lack of documentation. If your process prevents you from working with others—if your flows are spaghetti and notes indecipherable—you can’t produce the right outputs. 

As I already alluded, it’s exciting that nobody has defined best practices yet—which means you get to. If you have ideas to share, others want to hear them. Seriously. Right now, this profession feels like a high school dance where everyone is milling about the edges waiting for a brave person to step out. I promise you, you do not need to know how to dance. Contribute to the Voiceflow blog. We can teach you. Some of our best contributors have been people who didn’t think they were ready.

In fact, I urge you to make that a resolution for next year: share more, write more, and help us all legitimize this industry.

A tailwind next year: greater investment in conversation design

One silver lining to all that’s happening with the economy and layoffs is that every company is interested in automation, and doing more with less. So, while they are divesting in lots of areas, it probably means they're going to increase their investment in conversational AI. 

A headwind next year: difficulty calculating our impact

Designers have long struggled to justify their return on investment in tangible, immediate terms. Direct response marketers running digital ads have a much easier time. They can show the cost of acquisition immediately, and can turn it on and off like a tap. Whereas it’s much more difficult to define the benefits of a new brand identity, or the perceptual shift from full live agent support to partially conversational assistant-guided support. 

Next year, I bet a lot of companies are going to challenge conversation design projects by asking about the return, and lots of designers will struggle to respond. Their projects may lose out to surer short-term bets. Yet at the same time, I do hear lots of executives talking about investing for the long-term despite the environment. So, maybe the tailwind will overpower the headwind.

Voiceflow surpassed 100,000 users

This year more than 100,000 people used Voiceflow. We doubled our team size and the number of customers and I am elated. Seriously, thank you for those votes of confidence. We will repay it in innovation. 

One thing I’d take with me: you and the entire community

As always, thank you for being a member of our growing Voiceflow community. We truly have the best customers and we feel incredibly fortunate to have users who are as passionate about what we're building as we are. Without your impromptu product feedback, bug reporting, and forum posting, we wouldn't be anywhere near where we are today. So again, thank you. Our mission at Voiceflow is to democratize the creation of AI Assistants and together, we will.

See you in '23!