The voicebot ecosystem is growing immensely — and amazing opportunities abound. Reading a recent post by Alon Bonder, and realizing the main subject of conversation for product managers, startups, and developers is voice-tech, I figured out some points to help you focus on building the right product for what’s coming next. Basically, the mobile apps ecosystem we saw growing 10 years ago is making a return, but this time it is all about…voice.

Remember when?

In the beginning, before the mobile apps ecosystem rose in popularity, problems weren’t as clear as they are today. Specific iOS apps had memory problems; the UI was too simple; the development platforms were horrid (or nonexistent); there weren’t enough solutions for mobile app marketing, acquisition, and attribution; and the competition featured apps alongside thousands of farting, semi-funny, and non-valuable apps.

But as the ecosystem evolved and matured it granted new options to individuals and startups, who went ahead and made an “app for that.” These might be heaven-sent or perhaps simply tell you if something is Not a Hotdog.

Back to the future

Developers these days are struggling with incomplete voice platforms: Alexa, Cortana, Siri… you name it. Even if an amazing voice app is built, the ecosystem isn’t necessarily ready for prime time, or the full funnel: Develop → acquire → on-board → retain → make money.

The voice ecosystem is missing essential tools — available in the mobile apps ecosystem — to conduct appropriate analytics and measurement, marketing attribution, A/B testing, deep-linking for improved acquisition and re-engagement, and so on.

There are development solutions available for basic voice products — thanks to APIs, frameworks, and AI tools — but these are basic and, in most cases, only allow you to build a proof of concept without acquiring real users.

That leads us to a series of problems and opportunities.

Voice-tech problems and opportunities

1. Discovery

Discovery: Building a voice app is the first logical step, but finding an audience is the first difficult step.

How should developers distribute their apps? Try to tell Alexa to order you a cab, ask Cortana to transfer $100 to a friend, or ask either to find you a good payment skill/app. Voice Ad networks, affiliations, and more are challenging. Any personal assistant or voice interface is available by chat and can disrupt word-of-mouth as we know it.

2. Acquisition

Discovery isn’t that good now, so we need to promote our skill on Facebook, Google, maybe Twitter. Simple enough, but don’t we need a skill URL? How about the ability to enable the skill from the ad (like app downloads/installs), or track behavior after an ad was clicked? Unfortunately, that’s not available just yet. Appsflyer, for example, has been providing amazing attribution for the mobile apps ecosystem, but we need a similar solution for the voice ecosystem.

3. Onboarding

Happy times, a new user connected to your voice app — but will they use the skill?

To provide a smooth and practical onboarding experience, we must develop a proper, flexible, AI/ML-based tool that will talk the user through the experience to help them achieve their goals. Think WalkMe but with voice…maybe TalkMe? This can be combined with attribution, so the talk-through may be personalized for the individual user and help you find their preferences, age, and gender. Of course, you’ll need proper analytics tools like GA or MixPanel (or Voicelabs), and a real-time content platform to analyze, improve, and test your onboarding funnel.

4. Optimization

What’s a common known with the mobile apps funnel these days is non-existent for voice. We’re missing a tool for in-depth analysis that would grant us insights to understand, change, test, and optimize the experience of the new skill-enabled user — kind of like an Apptimize, but for voice. Also consider the conversion optimization ecosystem (Qualaroo, Unbounce) and the amazing possibilities voice apps are opening.

5. Retention

Did you know voice app retention is around 3 percent after seven days? In other words, 97 out of 100 users will not use your voice app after seven days. Crazy churn! Trust is one of the top reasons for churn, or the lack of trust. To build trust, the AI must understand how users perceive the app’s voice, tone, and tempo. Voice analytics will truly help us understand the bot and the user.

Push notifications must also be adopted by the voice ecosystem to help in the retention department, as Appboy or Urbanairship have been doing for mobile. Alexa‘s approach is a good first step but should be improved to include real-life communication between people. For example, if your friend wants to call you with an update about the game tonight, they will call you and not send a red-colored LED. That’s a given.

6. Virality

How do users bring new users to your voice app? A click on that Facebook or Twitter ad won’t do…remember, people don’t click. But you can ask “Alexa, please share Uber with Dan.” Social sharing is difficult when done by voice only, so we must create a “voice-sharing experience.” A tool to share my actual Cortana email experience can make this personalized and trustworthy, so we can listen and understand the value of voice.

How about social proof, such as “Rate our app!” and “please add your 5-star review so new users will think our voice app is amazing.” Alexa is there…Cortana and others aren’t yet. And the ecosystem? Build an API allowing users to share amazing skills and developers to easily track them to understand how viral the voice app is. And let’s not forget the right tools to ask for feedback…by voice. That’s difficult, as you don’t want a robot to interrupt while it’s helping you navigate.

The next steps

What’s next? Invite three friends (tech, biz dev, product) for a brainstorming session with beer and snacks, and read this post again. Then list thee top products from the mobile apps ecosystem and how they could evolve to the voice ecosystem. Then…build one!

Ariel Kedem is the VP of Products at Knowmail, an AI messaging system.

Above: The Machine Intelligence Landscape This article is part of our Artificial Intelligence series. You can download a high-resolution version of the landscape featuring 288 companies by clicking the image.