Not only do more people use messaging services than social networks, they spend more time using these services than they do on social networks and browsing the web, combined. Messaging is in its heyday. The future of technology is in conversation.

At the same time, breakthroughs in deep learning and reinforcement learning have brought artificial intelligence and natural language processing (NLP) to a whole new level. This goes beyond rebranding terms like “data mining” and “machine learning” as artificial intelligence. We’re now starting to see actual speech recognition, natural language understanding, and conversational interfaces that are both intuitive and useful on a human level.

The exponential growth in the number of programmatic services and interfaces not only put API back on VC buzzword hotlists, it also lets these conversational apps hail a car, order pizza, or even conversationally pair the best wine with your chicken parmesan. This shift is so significant that the question is no longer whether messaging will be the next big thing or not, but how this powerful platform can most effectively harness its potential.

There are two main trends in the instant messaging realm today: the chatbot and the personal assistant. While they are similar, each has its own unique features and applications. Put another way, they are two sides of the same software coin.

Dumb devices, clever bots

Some messaging apps are being considered “eyeball billionaires,” thanks to gigantic user bases, and brands that want to market to those eyeballs are taking notice.

The screen — or any terminal access point to a device that a user keeps nearby — is king. When brands get that close to the user, it opens up new marketing and revenue potentials. App fatigue is real, so users spend the majority of their time using simpler software — namely, messaging apps — and despite being rather limited, these apps do the job of connecting the user with friends and family. Brands are understandably eager to join these inside contacts that the user engages with.

We’ve seen a lot of creative experiments throughout 2016, but they all seem to converge toward having business accounts live seamlessly among human accounts, with an ability to send notifications, broadcast messages, and turn the chat thread into a user interface — Consider Kik, Telegram, Facebook, and Line, with Viber due to follow. Critics point out that among the hundreds of thousands of chatbots already out there, it’s hard to find any that are useful, but that’s how evolution works. These bots will breed, mutate, and adapt until developers get it right.

For this Darwinian process to work, the chatbots need attention not just from startups and venture capitalists, but from the messenger audience as well. This is not happening yet.

There are over 33,000 bots for Facebook, 19,000 bots for Kik, over 10,000 for Line, and hundreds of thousands from ManyChat and ChatFuel (most of which operate on Telegram). With this many bots out in the wild, how will users ever know what’s good? The bot discovery tools that messaging apps provide are no better than a contact list, and while most of the audience is in the habit of Googling things, these chatbots are widely overlooked by Google’s data crawlers.

There are curated catalogs that call to mind the early days of the web, but just as it was with websites, curation only works until you have several thousands of objects. It doesn’t work when there are millions of them, and chatbots are pushing this limit. BotList says it has more than 11,000 bots, and a quick look in the Google index shows that BotPages has less than 2,450. StoreBot leads the pack with just over 31,000.

Clever devices, dumb APIs

Personal assistants are on the rise, so it’s necessary to look at them as well, albeit a little differently. People use about five different apps and a couple of messaging apps, on average, but it is only a matter of time before human attention fades away from the smartphone. Soon our favorite toys will include smart home devices, smart cars, and even augmented reality systems.

When this happens, the smartphone’s monopoly on our attention will instead shift to something that brings all these focal points together. This suggests services like Siri, Cortana, Amazon Alexa, Google Assistant, Facebook M, and Samsung Viv. The dream scenario is that these personal assistants (or something like them) will help the user with anything and everything. They might answer daily correspondence, help you decide what to buy, or perhaps most importantly, choose where to buy those things.

Thanks to speech recognition and natural language understanding, each of the personal assistants above is already pretty good at understanding user intent. Fulfillment is a bit trickier, which is clear from the patent on evolving cognitive architecture filed by Viv. Viv is now Samsung’s Viv, embroiled in the first of many epic patent battles we’re likely to see.

The programmatic personal assistant doesn’t need all those clever chatbots and conversational interfaces; it just needs a decent service platform. For a human assistant, it’s Google, a phone number to call, and somebody to speak to on the other end of the line. For a digital assistant, it’s a lot of good APIs mapped to accurate ontologies, so that any user intent would be routed to and fulfilled by an additional third-party service.

Unsurprisingly, the big companies have service platforms: Alexa Skill Kit, Apple Siri Kit, Cortana Actions, and Google Now Integrations, among others. These are crowdsourced and curated, as are general API aggregation products like Mashape Marketplace, 21.co Marketplace, and IBM Harmony. They require human effort to stay current with the services that they integrate with, posing a limit to the number of services available.

For the aforementioned service platforms, that limit is hundreds or even thousands of services. However, API management is heating up, and the number of services will continue to grow. Some of the API management companies already host just under 300,000 APIs, and the recent Apigee acquisition by Google is likely to make things even more interesting.

It seems here the path to the world domination also has a discovery bottleneck, this time discovery for machines. Software cooperation currently requires a lot of human effort; you need manual work to discover an API, write the integration code, and pay for the service.

This town is big enough for both

There are clear signs that both chatbots and personal assistants will thrive.

Plenty of verticals can’t meet user expectations without a conversation (consider customer support, medicine, consulting, and legal services). At some level, conventional human conversation will remain the gold standard, but that doesn’t mean that many of the entry-level questions and concerns can’t be handled by software. That software ought to be human-like, and so we see slick-tongued chatbots beginning to emerge.

Brands have invested significantly to position themselves in our hearts and minds, and they will definitely continue to do so in order to avoid being aggregated through sophisticated (but heartless) real-time bidding systems.

Meanwhile, service aggregation already thrives in a number of verticals like transportation, lodging, and food, with more to come. Economically ineffective intermediation was replaced by centralized aggregation, dramatically lowering prices for the end users. But this leaves merchants and service providers feeling out of control (not that they had more control before).

Those two paradigms may coexist in a number of ways, whether it’s with a communication intent in your favorite personal assistant or a personal assistant bot in your messenger. They are two sides of the same coin.

Why not bet on both?