Back in December, when Slack integrations launched, I wrote Clippy’s Revenge about the potential of “smart messaging” to become a new platform. Since then, big players have done much to nurture that possibility  —  as if on some secret, jointly agreed-upon master schedule.

  • March: Microsoft released a bot framework at BUILD
  • April: Facebook opened up its Messenger platform at F8, and Telegram announced a prize for bot developers
  • May: Google announced its own Allo Messenger and voice-enabled home speaker at I/O, and Amazon made the sneakily successful Alexa accessible in a browser, without Echo hardware
  • June: Today at WWDC, Apple finally opened up iMessage to third-party integrations and announced the Siri SDK

This looks like the battleground of the next tech war, and all eyes are on Apple this week. But how much of the bot craze is hype, and what’s worth paying attention to? The frenetic energy around this emerging ecosystem is well-placed but often confusing.

After meeting with more than 50 founders in this space, I’d like to offer a structured explanation of the emerging conversational economy and to propose some opportunities (for both big companies and startups).

The conversational economy encompasses the growth of:

1) messaging applications, 2) voice-controlled computing, 3) bots and services that sit  —  or just start  —  within messaging apps / voice-controlled hardware, and 4) enabling, picks-and-shovels products.

The below will get updated with links as I go.

Drivers: What’s causing the bot craze?

Most people in Silicon Valley who are talking about the “conversational opportunity” are focused on the current crop of enabling technical drivers  –  what I call the “4 M’s” of next-generation software.

Rather than focusing on these (plenty has been written about them already) I’m going to emphasize the less explored yet more important economic, cultural, and ecosystem factors. Enabling technologies mean that the conversational economy can emerge, not that it will.

But, as context, the 4 M’s are:

  • Moore’s Law  — cheaper compute enables processing-heavy applications that were previously prohibitively expensive
  • Multi-tenancy  — on-demand public cloud providers dramatically reduce the cost and complexity of building and scaling a new service
  • Machine learning  —  advances (particularly deep learning algorithms), have proven unreasonably effective at solving problems that are very difficult to write traditional software programs to solve — for example, natural language processing  —  underlying technology for smart “agent” services
  • Mobile data — collection of rich, real-time data on smartphones provides both automatic user context and enough data for ML to be effective

Economic and cultural factors

Pervasive communications. The average teenager today communicates digitally 10x as much as the previous generation, and 100x as much as the generation before that. More than a decade ago, academics such as Thurlow described a “communication imperative” –  human beings are driven to maximize their communication volume and satisfaction. More recently, researchers have described it as a compulsion.

The mobilization of over-the-top messaging has fully enabled this compulsion  — it makes communication practically free once a person is connected to the internet (removing volume constraints) and enables fast, media-rich innovation in expression (e.g. images, video, emoji, GIFs, filters). Messaging  — as compared to voice or video  —  is fast, low-effort, asynchronous, private, and relatively unobtrusive.

We are just now seeing the first adults who grew up able to express their communication imperative in a pervasive way. They spend most of their computing time conversing. The fact that we spend so much time communicating is itself an argument that our communications platform should be our mobile platform.

(In Clippy’s Revenge, I provide some personal data about time spent in a messaging interface).

Consumerization of software expectations. The average person interacts with >10x the variety of software they did a decade ago. Consumer internet software companies, whose lifeblood is organic product adoption and engagement, continuously push the envelope in pursuit of eyeballs (and now, app-opens). They constantly redesign their apps to be ever-faster and more intuitive to use. Mobile distribution has allowed that innovation to reach consumers at a daily, weekly, and monthly cadence. As a result, consumers’ expectations around software UX continue to rise  —  they are looking for delight.

However, most legacy enterprise software companies fail to deliver. This is due to other heavy requirements (availability, performance, integrity, security), customer lock-in reducing pressure for vendors to change, buyers often being disconnected from the end-users, and technical debt. Of course, big company stagnation and Hooli-like management issues play a role as well. Others have written about this in sad but hilarious detail.

Legacy enterprise software isn’t keeping pace with rising consumer expectations  — it’s even a recruiting issue with younger workers. Many developers see integration with pleasant, fast-moving, highly engaging messaging platforms that feel like consumer software (Slack) as a potential outlet for consumer frustrations with rotting systems.

Mobile in the East >>> mobile in the West. The structure of the mobile ecosystem has actually evolved both faster and differently in markets where it was the first mainstream computing platform. Rather than being just a “new UX craze,” rich messaging applications are the primary starting point for humans’ digital lives in parts of the world.

For some 700 million Chinese people, WeChat is the dominant digital identity service, contact book, communications channel to both other humans and to businesses, and channel for commerce and support. It serves as users’ B2C and P2P payments rails. It is the preferred search engine for real-world, mobile information, including location and image and sound-based search. Users can go through Didi / Uber driver onboarding flows within WeChat, buy 20-cent melons from a street stand in a rural city by scanning a QR code within WeChat, or interact with 10,000+ other “official accounts” all within WeChat  —  a tiny 34MB app. Connie Chan and Mary Meeker both cover this phenomenon extensively.

The consumer internet is a partially (and increasingly) global market  — there is cross-pollination of actual software, of software idioms, and of ambition.

Within the confines of a single “app,” WeChat has built a better app store than Apple, serving a much broader set of use-cases. When U.S.-based developers, product people, and founders interact with WeChat, they inevitably see the potential as transferable. WeChat-envy is itself a driver.

Penetration of smartphones in developing economies. I’ll talk about developed-economy “app fatigue” later on, but the barrier to downloading and using new mobile apps is even higher in developing countries where the cost of mobile internet service represents a much more significant percentage of per-capita income. Google and Facebook are investing in developing-economy mobile data access with initiatives such as Loon and Free Basics. Startups such as LotusFlare want to offer similar services for other app developers.

Restrictions on communications to consumers via SMS are also weaker in other parts of the world than in the U.S. Many consumers would actually rather receive marketing via SMS than email, and SMS has magnitudes better open and response rates. In addition, the mobile OS, browser, hardware and app store ecosystem is often much more fragmented in developing countries. This raises the developer burden for native apps (Just imagine the test-and-deploy infrastructure!).

Being able to deliver services without requiring a full OS-specific app download will have an even stronger value prop in developing economies. The conversational economy is also not limited to mobile apps  —  services over SMS (such as mobile banking) are much more popular in emerging economies.

Transactional mobile services. Less than a decade after the launch of the first smartphone, massive mobile services with transactional business models (most obviously Uber) have emerged. Unlike the enterprise software players, Uber is less concerned that it is sitting in another player’s interface  — it wants its users to call Ubers from whatever interface they want, as long as they take more rides.

What other real-world services can exist like this? Mobile-enabled food and grocery delivery, self-storage, personal finance, commerce and real-world marketplaces don’t have business models directly dependent on app opens and eyeballs. In contrast, entertainment apps like Netflix do. While the quality of a native mobile experience may be an advantage for these transactional businesses, reaching new customers is usually a higher priority than owning the customer experience.

For a company that charges the user for a service outside of the phone (rather than making money off of engagement in an app) messaging platforms represent upside, potentially bringing them lower-friction new users and incremental revenue.

Monitoring → automation of knowledge work. When work product and work communication are digitized (as in messaging systems such as Slack, and communications-rich productivity applications such as Figma and Quip), a valuable new dataset is being collected.

Facebook serves up the information it thinks you want to see, because it has thousands (millions?) of interactions per user showing what you’ve clicked, read, and said before. It’s not hard to picture FB automatically posting “Happy Birthday!” to my close friends’ profiles for me. “Enterprise social networking” companies such as Chatter, Yammer, and Jive never achieved the kind of engagement that Slack is achieving today.

Every large company in the world is interested in knowledge-worker automation. Communication is a huge part of our work. New messaging platforms and communications-native applications offer richer (if unstructured) passive data collection about how we work than we’ve ever had before. This data collection is a necessary foundation to automation and augmentation.

Messaging “magic curtain” could balance business / human needs. I introduced the “communication imperative” earlier  — one of the most interesting things about it is what I call the “magic curtain” of messaging. Telephone interactive voice response (IVR) systems, with their robotic voices easily identifiable as software, are almost universally disliked. The most common IVR interaction is “0 for Operator” (or I’d bet, as frustration levels rise when callers can’t reach a human operator, “F for F*** You”). Yet every large company tries to de-personalize support  — because personally and instantaneously serving customers 24/7 with human agents is expensive.

Text offers a channel through which, with sophisticated enough software, it can be very hard to tell what’s real  —  it’s the Turing Test, everywhere. Brian Christian’s The Most Human Human is a fascinating foray into this idea. Messaging could carry context, remove the burden of IVR or web/app GUI navigation from the customer, and enable iterative discovery of needs. It also enables scalable “human” outreach, as in inside sales organizations.

The “magic curtain” of messaging may enable software to serve customers efficiently, making them happy at an acceptable cost to businesses. Commerce, support, and sales are the obvious “communications” service roles that you could pull the curtain over.

Everyone but Apple wants a different ecosystem. The biggest technology players have data, social graph, and distribution advantages (as evidenced by the amazing growth of Facebook Messenger). In the case of Google and Apple, they control as much of the consumer mobile experience as they choose to. But chinks are appearing in the seemingly impenetrable armor of these giants precisely because they are vying so violently against one another for domination.

App fatigue. Businesses and developers need to reach consumers on mobile, but it’s become extraordinarily hard to do so with new services. Consumers spend most of their time in apps created by Apple, Facebook, and Google, and the majority of smartphone owners download zero new apps per month.

Daily and weekly use-cases (perhaps monthly with a sharp enough pain) dominate the apps that get downloaded and used. But what about all of the software we consumers and workers use less often, the brands we may interact with only once every few months or spontaneously when we don’t have a fast connection? Today, the only outlet is mobile web  —  there’s an enormous gap in capabilities. I’ll dive more deeply into this (and Instant Apps and Subscriptions) in a future “Outcomes and alternatives” post.

Native mobile apps are hard to build. Building, maintaining, and refreshing mobile apps on two platforms (often alongside a web service) is hard, expensive, and out of reach for many. Messaging platforms do a better job balancing development effort with value to the customer by providing rich common services (payments, identity, context, social graph). Beyond conversational interfaces potentially being cross-platform, the UI surface is also dramatically reduced, especially at the outset. No longer do companies need to start with compelling visual design, app user boarding flows, and slick gesture interactions.

Developers and businesses can start with a “thinner wedge,” use shared services, experiment manually (sending human-crafted messages and content), and progressively automate and expand into rich interactions, rather than the somewhat all-or-nothing experience of building native apps.

Poor integration leads to poor experiences. The “native app” container and tight control of the App Store by Apple has led to more polished experiences on some dimensions, and better security. Apps may break less when they’ve gone through an approvals team  —  but experiences are higher-friction in other ways. Why should users have to create new accounts just to try any new mobile-native functionality? Enter payments information every time they shop? Manage granular permissions for every app individually? Copy-paste passwords?

The controls that operating system players enforce are choking their ecosystems: lengthy, unpredictable app-store approval processes and heavy development burden, stagnation of shared services, and locked-down environments that hamper consumer experiences. This explains much of the explosion of developer interest in building conversational applications.

Image credit: Matt Schlicht, who both created this group and collected the data for me

Apple’s rivals pin hopes on “new computing devices.” Dissatisfied with the mobile status quo, Amazon’s Alexa Echo and Microsoft’s Cortana both represent bets that there will be new centers of gravity for consumer computing. These tech players are starting with contexts in which our small-screened phones are challenged, or where we can deliver something better  —  when we’re driving our cars or hanging out at home. Consumer and developer interest in Alexa has caught many by surprise. Seeing the potential, and its alignment with Google’s self-declared advantage in A.I. to power conversational voice experiences, Google has also acquired and announced its own projects: Nest, OnHub, and Home.

Apple’s rivals are hungry for new computing platforms, and the home and the car happen to be contexts where mobile is challenged  —  but hands-free voice interfaces make sense.

Strong parallels in B2B software ecosystem. As alluded to above, the snail’s pace of UX evolution in enterprise software has led to extremely low engagement. The vast majority of enterprise applications, seven years later, do not have first-class mobile interfaces  — for example, a Fortune 100 aerospace/defense company recently described to me their workaround for a core inventory application lacking a “mobile mode:” They are rendering virtual desktop views of their core inventory application on shop-floor iPads, but the slow VDI performance and clunky workflows are driving their workers crazy.

The stagnation of existing software and dramatically increased consumer software procurement, combined with the Cambrian explosion of end-user friendly SaaS built on public cloud resources, has created an architecture of fragmented data trapped in small, poorly integrated applications.

Growing companies such MuleSoft and Segment already offer approaches to the SaaS integration problem, but Slack is pitching itself as not only the integration answer, but also as the “user-native” answer.

Tailwinds abound for the conversational economy. The technical triggers are the reason this shift in mobile software can happen. But just because it can happen, doesn’t mean it will  —  the economic, cultural, and especially, ecosystem triggers that people have been less focused on are actually more important.

The state of the mobile ecosystem (OS players currently failing to enable businesses and developers) is as much a driver for the conversational economy as the technical, economic, and cultural factors.

Apple’s initial announcements around iMessage and Siri today will only accelerate interest and will raise many new questions. What will the iMessage bot approval process look like? Will iMessage come to Android?

One thing is clear  —  this is the new tech battleground, because the big players have made it so.

Stay tuned for Part 2: The Conversational Economy, Drags (And Why the App Economy May Still Rule)

A version of this post appeared on Medium.