Are chatbots an evolution or a revolution?

Chatbots are a hot topic in tech — positioned at the intersection of many macro tech trends like the rise of artificial intelligence, messaging apps surpassing social apps in users, and enterprises struggling with customer acquisition and engagement on mobile. Descriptions boldly range from the next mobile runtime to Google-killers. Investment dollars are gushing into the leading messaging apps, and all of the tech giants have announced significant new initiatives. If you have ever used Amazon Alexa, you can see what the hype is about.

But will chatbots be as revolutionary as the hype would have it? The evidence suggests otherwise.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":2042912,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,","session":"D"}']

Lessons from China and India: ubiquity and simplicity

To predict how chatbots will evolve, we can look to Asia, a region that has led the West in this space. We’ll use WeChat in China and SMS-based messaging in India as examples.

WeChat is often cited as the poster child for chatbots. With 500m+ monthly active users and $7 ARPU, it is definitely worthy of all the attention. Much has been written about WeChat’s success, though I would argue most of it is not relevant. For example, much of its functionality is through a browser interface rather than a chat interface — hardly a model for the promise of automated chatbots. WeChat was launched four years ago, when the mobile ecosystem was still nascent. Modern solutions must navigate a mature market with entrenched competitors.

On a more positive note, the monetization model and the online to offline (O2O) integration are the most compelling learnings from WeChat. Monetization is focused on payments rather than advertising. Messaging apps occupy a highly personal space, and advertising is ill-suited in this space without risking user backlash. Integrated payments, however, are an important additive part of the user workflow. O2O integration allows WeChat to have a ubiquitous presence in all aspects of one’s interaction with a brand. WeChat’s integration with in-store QR codes facilitates user acquisition, and the WeChat payment platform allows for in-store payments.

Application-to-person (A2P) SMS messaging in India may be a more relevant precedent for today’s chatbots. In India, A2P messages are a ubiquitous aspect of daily life. Users receive more than a dozen messages a day, and more than 10 billion messages are sent each month across India. This market has evolved in recent years from notification-oriented messages (‘you completed an ATM withdrawal’), to marketing-oriented messages (SMS, not email, is the dominant digital channel), to customer service-oriented messages (‘your order will be delivered today’), and now interaction-oriented messaging (‘tell me my last 3 transactions’).

The success of messaging in India is highly instructive for what can be successful in the West, but not without its challenges. The most successful use cases were lightweight — useful information with little cognitive load given the 160-character limit for SMS. Also, transactional use cases did not require user consent. Solving this user acquisition challenge significantly accelerated adoption. It also carried risks as brands blurred the lines between transactional and marketing content. This led to significant user backlash and ultimately onerous regulations to restrict unwanted marketing SMS. Lastly, SMS was a ubiquitous platform. From this experience, there are clear learnings around simplicity, platform ubiquity, and user acquisition. It is also a cautionary tale for platform providers who must self-regulate how and when brands can send marketing messages to their users.

Success in the West: evolution, not revolution

Reflecting on these experiences, there are a few takeaways that argue for a more evolutionary approach for chatbots in the West:

• Early use cases are likely to be quite simple.
• Rich interaction experiences (e.g., buttons, menus) are required to drive complex use cases.
• Cross-platform support is critical for widespread adoption, and there are already more than a dozen relevant platforms, including Messenger, WhatsApp, Alexa, iMessage, Skype, Slack, SMS, and email.
• Platforms must tread carefully on user acquisition, ensuring it is not too onerous as to hamper growth while ensuring users are not inundated with undesired messages.
• Platforms must self-regulate to resist the pull to make chatbots another marketing channel.
• Payment integration will be a key enabler for consumer monetization.

Some of these trends are already taking hold. Amazon Alexa has done well commercially, though the capabilities are relatively limited and the A.I.is largely limited to voice recognition. Facebook, an early leader, has focused much of its effort on being thoughtful about user acquisition for brands and has also been proactive in its payment integration strategy. Facebook is also regularly rolling out richer tools for building more engaging bots. Apple’s recently announced changes to iMessage are similarly related to the critical UI components for building chatbots.

Opportunities

Most of the early hype around chatbots has been for consumer applications. I believe most of these investments miss the mark and will accelerate the inevitable disillusionment phase of the market.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":2042912,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,","session":"D"}']

We often hear about shopping, concierge, and travel service chatbots — much of what is referred to as conversational commerce. Building an A.I.-enabled solution for these use cases is complex given the open-ended nature of the problem. Also, the business model is challenging. As stand-alone apps they suffer from typical discovery challenges of mobile apps. Trying to ride the rails of one of the major messaging platforms is fraught with commercial risks. I do believe there will be exciting consumer-focused use cases, though I predict that they will mostly be complementary to traditional apps rather than apps in their own right.

Companies that focus on enterprise use cases around productivity or workflow (like x.ai and Talla) are building compelling solutions for productivity enhancement and task management. Workflow-focused companies like Drift and DigitalGenius are helping manage real-time customer conversations by deploying A.I.-enabled automation of the agent response. While these problems are complex, they are relatively narrowly defined and hence more solvable. The business models are also simpler because the botmaker can directly charge the enterprise.

The largest opportunity for chat interfaces may be to integrate them into the tools we use on a daily basis. Companies like Troops and Kasisto are tackling straightforward use cases like updating a CRM or getting information from one’s bank, where a chat interface provides a faster and simpler experience than today’s need for a dedicated app. Many solutions can be built through straightforward integration with the APIs incumbent software providers already offer, which require only limited A.I. capabilities.

Despite the rapid progress, there has been limited focus on cross platform support and integration of chatbots, an area critical for driving ubiquitous adoption. Twilio recently announced cross-platform support, and startups like Smooch.io, Gupshup.io, Kore, and Chatfuel are developing solutions. But these are challenging problems that require a higher level of investment. Across each platform, there are unique models and varying levels of support for functionality such as:

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":2042912,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,","session":"D"}']

• User identity and authentication
• Context management (e.g., chat history across platforms)
• Platform preferences (e.g., restrictions on message sending)
• Content types (e.g., images, buttons, menus)
• Payment integration
• Bot testing and deployment

Cross-platform solutions that minimize the burdens on the bot developer will form the basis of an entire ecosystem of development, integration, and operational tools for chatbots, and will be a key enabler of growth.

Chatbots have the potential to be the next great innovation in software development. Early use cases, however, are likely to be evolutionary enhancements to how we interact with existing software. But as users become more comfortable with these interfaces, as the ecosystem matures, and as A.I. capabilities continue to improve, chatbots may indeed lead to a revolution in the software industry.