🧙🏼 Deep dive: AI voice agents

Howdy, wizards.

Today, I’m doing a deep dive on voice agents. I’m genuinely curious if you guys like this format or not – just reply to this email with “useful” (add comments too if you want) if you enjoyed it.

The context: OpenAI finally rolled out Advanced Voice Mode and launched the Realtime API – unlocking a lot of new use cases for voice. AI-powered Siri is rolling out over the next months. Copilot also added voice last week. Google’s NotebookLM has gone completely viral since they added the Audio Overview feature. Voice is the hottest new modality of AI.

Here’s what’s covered:

• Where we are in the adoption curve for AI voice agents and what the early use cases have in common

• Who the top players are in the market

• Some cost considerations to keep in mind

Let’s dive in!

DARIO’S PICKS

1. AI voice agents: adoption & early use cases

Venture capital firm a16z has some amazing points on where we are in the adoption of voice agents, and what commonalities the early adopters have.

We’re currently in the early adopters phase, with voice agents gaining momentum and interest across industries.

• It started with things like appointment booking, call centers, outbound sales, and restaurant orders, and now has made its way into drive-thrus, back-office healthcare, logistics, debt collection, recruiting and coaching.

• As the early successes become clear, new use cases like high-skill interviews and front-office healthcare are arising.

• The early adopters rely heavily on calls to drive new business. Also, they have a high-volume of calls with a defined structure, and they’re low-risk.

  • Example: Telecom Italia (TIM) implemented a Google-powered voice agent to address many customer calls, increasing efficiency by 20%.

‎ Why it matters‎ ‎ With developers now having the tools to build real-time voice apps, there’ll be a ton more use cases in the coming months. The enthusiasm is high and the threshold for building is lower than ever. As the we shift from early adopters to the early majority, though, we’ll see a lot of innovations either take off or fall flat.

TOGETHER WITH DRAFTBOARD

DARIO’S PICKS

2. The innovators of AI voice agents

a16z also released a thesis for AI voice agents back in May, which included a market map of both B2B and B2C players.

The B2B innovators and early adopters showcased here are in many cases businesses automating repetitive or high-volume calls, as we saw earlier.

• An example is Goodcall, which provides businesses with an AI that can automatically answer their customer’s calls, handles requests like scheduling appointments, and capture lead information.

The B2C use cases are more about learning and companionship.

• Apart from ChatGPT, I think the most broadly used AI voice agent right now is probably Character.ai, which falls under companionship. They have 20m+ monthly users and has had 107 million “character calls” on their platform in the last 4 months. Hot tip: if you’re curious what the hype is about and want to find some cool characters to call – you need to check out my new site BotFlix.ai.

• Speak, an AI-powered language tutor, was featured as an early tester of OpenAI’s Realtime API. Check out the demo (second video) to see how they applied speech-to-speech for real-time learning.

‎ Why it matters‎ ‎ Knowing who the early adopters of AI voice agents are enables investors to spot growth opportunities and builders to identify market gaps. Keep in mind that this view of the market is already a few months old, though, and rest assured it will be changing fast.

DARIO’S PICKS

3. The cost perspective of AI voice agents

As we’ve seen, AI voice agents are all the buzz, and even more so with OpenAI’s launch of the Realtime API this week. Let’s consider the cost aspect, with the example of call centers.

The Realtime API runs at $0.06 per minute for input and $0.24 per minute for output. So a relevant question: is it financially smart to replace human call center agents with AI at this point?

Peter Gostev, head of AI at Moonpig, made a cost breakdown which shows that using the Realtime API is pricier than outsourcing customer service to countries like the Philippines or India, where labor is affordable and English proficiency is high. If cost were the only factor, and prices were equal, most of us would probably prefer talking to a human anyway.

‎ Why it matters‎ ‎ Costs of voice agents are dropping quickly. With OpenAI’s Realtime voice API being cheaper than human call center agents in Western countries on its first week of launch, I don’t see this aspect being a hurdle to adoption for very long. Ensuring AI delivers a sufficiently high-quality experience for the business and the user, on the other hand, is a completely different beast. It will likely remain a challenge for a lot longer than costs.

RECOMMENDED READING

The pace of new stuff happening in AI is absolutely nuts. I want to remind you that there is no way I can cover it all myself. When I write this newsletter, I just try to think of the most impactful developments and trends serve only the key things to you in an easily digestible format. That’s why I’ve partnered with TLDR. They have a much broader coverage than I do, though they don’t go very deep into each topic. So if you want to make sure you never miss anything that’s going on, consider subscribing to TLDR’s free daily newsletter.

THAT’S ALL FOLKS!

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.