Bots are No Apps. Yet.
Bots — that can also be described as headless apps in a conversational messaging interface— have been hyped up as the next big thing.
But anyone who has spent any amount of time to find and use them on facebook, kik or any of the other messaging platforms will admit to three major issues.
Discovery. Engagement. Usability.
Discovery: Bots do not need a download, but are not going to be easier to discover than apps unless the messaging platforms produce a “google for bots”; one that is smarter, contextually intelligent and uses a more effective ranking and surfacing model. At the moment, finding or sharing bots is a terrible experience.
Engagement: Bots have no form or shape and while this can be a blessing, it is also a curse. What makes it great for dead simple use cases, also makes it unusable for most others.
Usability: Every app exists to address specific “jobs to be done”. When user interaction is simple (typically one-button), and the user intent is super-specific, bots can be a great substitute for apps. Not dissimilar to how we search for google from the browser toolbar or call Uber from Google Maps.
However, when the user needs to explore and iterate to make a decision, bots in their current incarnations simply do not cut it; they are way too rigid.
Try using Sephora on Kik, or 1800flowers on Facebook messenger.
It can work as a keynote demo but is likely to vex and frustrate real people trying to express needs and get to a solution.
A Framework to Understand Where Bots Fit
One approach may be to classify use cases by user intent (specific, exploratory) and user interaction complexity (simple, complex).
User intent = specific: User knows what she wants. e.g. calling a Uber.
User intent = exploratory: She has a general idea of what she wants but needs help in figuring out the specific answer. e.g. looking for the best Italian restaurant or the best hotel to stay tonight. This likely requires a dialog with multiple follow-ups.
User interaction = simple: User has to press a simple button or follow a set of sequential steps.
User interaction = complex: There is a lot of iteration and asynchronous navigation to collect varied data points, and go back and forth.
This 2*2 results in four major types of use cases.
Push Button: Call Taxi, Order standard pizza, Check-in etc. This is the sweet spot for bots. These are also typical and standard API usage scenarios.
Search and Dialog: User expresses needs usually through search or selecting from a list of options; User needs to evaluate the results and make a decision on what action to do, each of which may involve further follow-up interactions. e.g. find cheapest or nearest hotel tonight, find the fastest ride, book movie tickets etc. This gets into more tricky territory for bots and they perform poorly, at least for now. In the medium to long term, this quadrant may drive the evolution of a bot concierge (owned by the platform) who brokers across multiple suppliers to find the best service for the customer.
Structured and Linear: Order flowers, deposit checks, log a customer support request etc. Current bot implementations take the form of a wizard navigating the user through a hierarchal menu of questions and responses. As many have noted, current bots addressing this need are fairly primitive and risk alienating users accustomed to well fleshed out app alternatives.
Unstructured: Window shopping, Browsing books, Exploring new fashions etc.This may be the hardest area for bots to conquer as the user ideally wants a wide range of data points and is used to asynchronous navigation in apps and websites.
Bottom-line: Bots offer exciting possibilities but are no apps, let alone superseding them. They do not address discovery — the biggest challenge with apps — in a meaningful way. And they are not engaging or usable, but for the simplest of use cases. Mobile apps have their issues but they work and are liked by users. And Apple and Google are likely to be doing a lot more to address the gaps around discovery and intelligence.
In the short term (0-2 years), bots are likely to be used more like an API channel that can complement apps vs. a replacement. Also, expect to see a fair bit of human augmentation by businesses. There is a real demand for chatting with real humans who can help with customer support. (SeeHaptik Mobile for a good example). This may not scale as well as automation but could work better for users.
In the medium term (next 2–4 years), as bot platforms help developers create more engaging, intelligent and usable experiences, they have the potential to emerge as a viable substitute to apps for a broader variety of use cases. However, key capabilities — the NLP, machine learning from social and bot usage data, interaction wizards, range of data widgets, ease of integration to back-end business services, and ability to deal with non-standard requests — need to evolve substantially for this to become a reality.
In the long term, we may see the emergence of intelligent assistants who act as high value-add concierges and seek to curate and mediate across the internet. And work across apps, bots, websites and beyond. See more discussion here.
The future evolves in mysterious ways and is really a constantly morphing distribution of outcomes, each with varying probabilities and dependent on many variables.
It is subjective until it arrives.