Where’s the Killer Chatbot?

shutterstock_408097675-technode
Image courtesy TechNode

Let’s face it: Most of the chatbot experiences today are pretty wretched. They’re stilted, artificial and in some cases downright affected. Natural language processing is still in its infancy, and has a long way to go before sounding actually “natural.” Or truly understanding natural speech, for that matter.

This is due in part to the difficulty of designing a user interface around a conversation, which is non-hierarchical in nature. When talking to another person, the steps don’t always flow naturally from one to the other. This kind of design is also fundamentally different than either a mobile or web interface. Additionally, we have yet to develop a general-purpose AI which can accept a user’s open-ended input.

UIs are crafted to serve linear experiences

It is incumbent on chatbot creators, therefore, to pick out engaging patterns of interaction. Building on and around these will enable developers to create whole experiences that will delight the users.

So how do we work around the limitations of a conversational UI, knowing the above?

About the UI
Up until now, User Interfaces have been crafted for a linear experience, not a random one. In other words, after the user comes to the page, a specific sequence of events typically happen, at least in terms of ecommerce. First they search for an item or two. Those items are then added to the user’s cart. They enter payment information, check out and leave the site.

A chat based UI is completely different from either a web or mobile interface. One of the biggest stumbling blocks is that the customer can initiate the procedure in different places. Say they want to buy tickets for a movie. The customer can ask a bot “What’s playing around 8pm?” Another valid starting point can be “I want three tickets to Trolls at the Regal on Little Texas Lane and Congress.”

So we see that a big challenge for anyone wanting to design a chatbot is that the path a customer will use to reach their goal (in this case, to purchase tickets) is not known beforehand. The chatbot has to assist the user and provide the desired answers without needing a discussion to progress in a straight line.

AI is Not Yet Ready
The next big stumbling block for chatbot developers is that a true AI that works on a variety of inputs is still a long way off. AIs themselves are not especially new, but they are new to the consumer marketplace. One AI-like construct that bot creators use a lot is the Simple Linear Tree, which forces the user down a predetermined path. New AI routines might also be used, but these are not true AI. They simply match patterns against pre-programmed conditions, in an effort to determine a user’s intent.

What we think of as AI is not truly AI

Generally speaking, these work well enough when there are a finite set of ways a user can interact with a bot. But as developers are finding out, user input can be totally random. This leads to situations where a bot can get unexpected input that it can’t handle. So without better tools, a better AI, it’s all a matter of hunt and peck. Or worse, finding the linguistic needle in a haystack of possibilities.

The Solution: Modify, Publish, Iterate, Repeat
So how does a bot developer succeed with the limited tools they have? The best path is not already defined, given the variety of inputs. Neither the number of inputs nor their content is known. There has to be a quick, iterative path to successful completion, and it has to be low-cost as well. A developer needs to be aware of how their bots are responding to the inputs provided by the user. With this knowledge, they can then iterate on what is already there. Any blocks between the user and their goal need to be addressed.

Users must be able to easily understand your conversational UI

Experience has shown that the best tools for the iterative method are bot native. This means they are able to understand the complexity and nuance of a conversational interface, and are able to translate them into clear metrics. Conversely, it also means the user is not simply dumped into meaningless dialogues or dashboards.

Marketing teams can use these tools to pinpoint groups of similar users, then connect with them through personalized messages. Creative and editorial teams can use them to address messaging that may be off-brand or that doesn’t have the desired tone. Business leaders can use them to provide a detailed picture of their efforts without the use of an engineering team and a data scientist just to “run the numbers.”

It’s important to have a conversational UI that’s easy to understand. It’s also important to iterate quickly on this. Being able to do these things will assist business leaders to grow differentiated bot-native arms that can leverage the great power found behind the conversational interface.

8 Truly Helpful Chatbots

chatbots-chatboten

There’s a lot of hype around artificial intelligence today.  And chatbots are one of the primary ways the consumer is made aware of AI.

Chatbots are surprisingly easy to create.  And the deluge shows no signs of slowing down, thanks to the hype, the ease of creation, accelerating investments, and the rush of developers.  Over 225,000 bot developers have created more than 300,000 bots, reports Pandorabots, a leading chatbot development platform.

In this case, though, there is lots of supply and not so much demand.  The rush of supply seems largely due to the “cool factor” of bots, and is not driven by anyone asking their favorite brands to start using chatbots.

Chatbots do well when focused on one task

Dr. BJ Fogg is the founder of the Persuasive Tech Lab at Stanford University.  When looking at his model of behavior, we see that in order for any shift to be made in consumer behavior, any given chatbot needs to be easier to use than the method being replaced.

Unfortunately, they’re not there yet.  Chatbots are still clunky.  As of this writing, it’s still easier to call, text or email the company in question.

At this point chatbots seem to do well when they focus on making a single task easier.  There are a few that realize the promise of AI, and that are able have more reasonable conversations by using context.  At this point, though, many users will still have only bare-bones conversations with most bots.

The sheer number of bots out there presents users with a problem:  How is anyone going to know which ones are truly helpful?  Sure, some bots may well assist a company’s cost-cutting measures.  Still, there is a lot of work needed before any customer gets a measurably better experience.

Right now there are a very few chatbots that provide some real utility and actual value.  They provide an inkling of what an interaction with an actually intelligent bot might be.  You might notice a theme among the bots picked:  They’re more realistic when the user wants to ask one of a group of prearranged questions.  They’re at their best when the tasks they handle are specific and clearly defined.

So here’s a list of the eight most useful chatbots out there today, in no particular order:

1.  Howdy’s function is to help you run your meetings in Slack.  Its founder, Ben Brown, created it with an eye to customization, so that everyone can use its talents of digital assistance and automation.  Howdy will reach out to participants and ask them a group of questions to prepare for the meeting.  It automates information collection and distribution.  It makes meetings shorter by leveraging a shortcoming of human nature — Howdy can talk to many people at once, whereas humans can’t.  These answers are then delivered to the meeting leader, and all participants, if the leader chooses to do so.  It sure beats copying and pasting the same questions over and over again.

2.  HealthTap was launched on Messenger not too long ago.  It aims to make good healthcare easier to reach by more people.  Not only does it allow a user to get a referral from real doctors, but it also assesses results of tests.  All that, and you can ask it health questions as well.  If other users have asked similar things, then it will show you those results.  Of course, health information can be highly personal, so HealthTap is ideal for those people that have no inhibitions.  The company does state that all questions and identity information is kept confidential and anonymous, but the fact that the company may associate a personal Facebook account with certain medical conditions may be a turn-off for some that prefer to keep a tight lid on their health information.

3.  Sephora chatbot on Kik.  Although it currently exists only on the Kik chat platform, the Sephora chatbot makes it easy for consumers everywhere not only to shop for products, but to get beauty tips as well.  Acting like the best in-store assistant, it provides three prime attributes of great customer service:  It responds right away, it’s highly educated, and it’s always available.  If you’re a member of Kik, it’s like having access to a Sephora team member in your back pocket.

4.  X.ai provides Amy, a personal assistant with one function:  Amy takes on the task of scheduling meetings, so that you can focus on more important tasks.  X.ai has been so successful with Amy because she does only one thing.  But that one thing she does extremely well.  She’ll schedule meetings without effort, but won’t take notes or solve any other problems for you.  Dennis Mortensen, X.ai’s founder has made a huge bet on the thoughts that bots are terrific for any very small task.  Due to this laser focus, the bot needs to be just about perfect.  Amy is their first attempt, and she makes scheduling conversational and easy.

5.  Assist integrates with several on-demand services such as Uber, GrubHub, OpenTable and Lyft.  A user can communicate with Assist through Facebook message, a text or by using Assist itself.  They can then get a ride, book hotels, or order take-out.  Assist uses input from users to improve itself with every question.

6.  Ozlo is a more personal AI.  At this time available only for iOS, he helps the user find information on their phone both easier and faster.  As with other bots, engage Ozlo in a dialogue text conversation, and it will give you answers on Google Now-like cards.  Deeper links are also provided to the websites and apps where it found the information.  The bot is partnered with other websites and apps such as Foursquare, Yelp, OpenTable and Zagat so he can provide great recommendations.  Right now Ozlo is being trained to help people find drinks, eats and entertainment — things that we all do daily. You can get more information about Ozlo from Charles Jolley, one of its chief designers.

7.  Pana blends AI with real humans to create an always-on travel agent.  Planning a trip is made easier, simpler, more personalized and cheaper with Pana’s skills.  Pana can also be your personal concierge, providing vetted choices for places to go and food to eat.  Since booking travel these days is mostly a matter of searching the same criteria on many sites, Pana makes booking easier and less time consuming.

8.  Birdly is a chatbot that acts as a bridge between Slack and Salesforce, making business-related data accessible to entire teams.  Directly on Slack, one thing Birdly can do is manage expenses.  This makes things more efficient, as it recognizes data input from invoices or receipts, allowing users to spend less time on expense reports.  Birdly also provides analytics, efficient alerting and customer information in one place.

What comes after building your chatbot?

after-the-chatbot_huffingtonpostuk
Image courtesy Huffington Post UK

Way to go!  You’ve just finished building a chatbot for your own awesome purpose.

But what comes next?

It’s easy to think the hardest thing is just building your bot.  It’s no lie that it takes hard work and dedication.  You have to make many tough decisions about what platform you’re going to use, what you want the experience of your users to be — and that’s not even including the conversations.  After all that is worked out, though, you’re going to release it to the world-at-large.

Choose a metric and aim for success!

And this is where the really difficult work starts.

In order to help you along, here’s some tips to make sure your chatbot is received well.

Know when you’re a success

Each product has at least one measurable item which will identify its success, or lack thereof.  There are lots of metrics to choose from.  Amount of messages traded per session is a good one.  So is total amount of chats sent and received.  Two more have to do with users:  Number of unique users, and total amount of returning users.

How will you know when you’re a success?  Pick a metric or two and aim for those.

What’s your path to success?

So now you know what success looks like to you.  Now you have to create a plan to actually get there.  Success doesn’t drop into your lap.  People won’t ever find you if you don’t have a plan.  They never contact random Facebook pages in hopes there’s some neat chatbot they might be able to use.

How do you achieve victory?  You get there through the use of measurable and intentional steps.

Your target audience should form the core of your path to success.  Learn where they go, where they congregate.  Then go there yourself and spread the gospel.  You’ll know you’ve arrived when your first users begin to market your bot to their friends.  Word-of-mouth is still one of the most powerful marketing tools out there.

Data, Data and More Data

Delve into your data on a regular basis.  You’ve already defined how you know when you’ve crossed the threshold into “success” territory.  Start there, then take a look at any following complementary metrics.  Gather this data and graph it.  This will make it easier to spot trends.  If your metrics are slipping, ask yourself why.  The answers lie in your data:  You should be able to see if your conversation paths are presenting users with dead ends.  Perhaps there was some big event that prevented your users from using the bot.  Or maybe they’re simply bored with what content you’re offering.

At the same time, should you see your metrics heading upwards, don’t just sit on your laurels.  Ask why that is.  Replicate a winning ad campaign.  If you made a great Reddit thread, try posting again somewhere else.  Make continuous iterations on the effort that boost your metrics the most.

Periodically review chatbot conversations by hand

Remember what Arthur Ashe said:  “Success is a journey, not a destination. The doing is often more important than the outcome.”  He probably wasn’t talking about chatbots, but the saying is true for them nonetheless.

Keep your bot fresh

Once you’ve coded your chatbot, it’s tempting to move on to the Next Big Thing and forget about it.  That puts your hard work on the fast track to obscurity, though.  You need to keep adding relevant conversation paths — and make sure you use the bot’s existing voice!  That fresh content will do wonders to re-engage returning users.

It’s important to stay ahead of the curve, especially with regard to natural language processing.  Make sure your bot can discern between words found in the dictionary and those that might be typed in.  In the latter case, be sure the user is shunted to correct path.  One great way to find out which words your bot needs more training on is to review user conversations manually.  Not only will this method help with those peculiar words, but you’ll also discern other ways to improve the overall experience.  Little tweaks can go a long way.

At the same time, this method makes it easy to find rabbit holes.  Don’t get tricked into going inside.  If one user is having a particular problem, make sure other users are having the same issue.  If they aren’t, don’t waste your time on fixing it.  Your bot’s identity is not worth compromising over just one user.

Analyze It Again and Again

Review your data points and user conversations over and over.  Track everything humanly possible and look for trends.  Use that data to keep improving.  Always.  Keep doing these things for the full life cycle of your chatbot.

For you and your bot, ultimate victory might consist of having thousands of users.  It might mean you trade jillions of messages.  Regardless of which metrics you choose, there is no winning without an investment of time and a great deal of tenacity.  Hang on to your goals; keep measuring your progress so you’re ultimately successful.  Stay aware of the time as well.  No sane person is going to wait around for months, just to see if your bot does something cool.

Now quit reading and go build your chatbot!

Writing Chatbot Dialogue Worthy of John Grisham

chatbotdialogue_ell-blue-icon

We’ve come full circle — once again, our interface with computers is defined by words.  Instead of specific words and phrases, though, we can now use the full range of the English language.  The entire lexicon is our UI/UX.  Theoretically, at least.

Exact attention must be paid, therefore, to each phrase of an interaction with a bot, just as it was once paid to every pixel and icon in an app or website.  Many people are ready to admit their lack of experience when it comes to creating that website.  Yet when it comes to the dialogue of a chatbot, we’re all suddenly experts.  Haven’t we all been talking since the age of two?  We must know at least a few things about communication by now.

Sophistication and planning is required when writing chatbot dialogue

Of course we’re quite skilled at relating snappy anecdotes to friends.  But we’re not as deft when it comes to succinctly relating a long-form technical process to a stranger.  For this, quite a bit of sophistication and planning is needed.  You have to grab and retain the attention of a person who is not concerned about damaging your ego if they terminate your conversation early.

So how can you craft irresistible chatbot dialogue which will ensnare users like a finely honed app interface?  A good place to start is by asking these questions.

Who is your audience?

Think about your “perfect user” a lot.  A LOT.  Describe them as narrowly as possible.  Sure, anyone can potentially use your chatbot.  Yet few things equally appeal to Queens grandmothers and L.A. tweens.  Who do you think will comprise the majority of your audience?  This will advise you on how you should craft your chatbot dialogue.

What is the end goal?

In other words, what action do you want your audience to take?  It’s important to know what you want from the user, what you want them to do, before you start rambling on without a point.

You need to have a plan

Once you’ve addressed the above questions, you need to start thinking about a general outline for your chat experience.  This outline will become the framework for all your later conversations, so it’s absolutely essential to have it to guide you as you’re writing the specifics of your chats.  Why do you need to do it this way?  Because if you don’t have a framework, your conversation will be devoid a point, and thus the entire conversation will be both confusing and meandering.  Not only that, but your users may well feel they don’t have enough guideposts to get back home.  They might even simply give up.

One strategy that works well is to make note of all the top-level points you want to get across.  Make sure it’s a manageable number.  No more than 10, perhaps.  Be aware of their order, and make sure they flow logically one to the other.  Your conversations should flow as well.  Be sure to place your most important points at the top of the list.  It’s a mistake to assume the user will make it completely through your chat maze every time, regardless of how brilliantly you designed it.  Hook your users early, and you have the luxury of educating them later.

Employ Artificial Intelligence, not Artifice in Communication

One thing our brains excel at is communicating with other people.  It’s an incredibly important skill, shaped by many years of evolution.  Typically we take stock of a room before speaking out loud.  There is no room, however, when communicating with others through a computer.  Thus many of us use a completely separate part of our brains to do so.  If someone is at a dinner party and is going to tell an off-color joke, they’ll usually be aware of who else might be present.  Most people will consider how those at the table may take the joke.

Think of your “ideal user” when writing chatbot text

But there is no context with online conversations.  There is no “room” to take stock of.  A cursory scan of any online comment section will suggest that most people don’t extend the same courtesy as they do in their real-life conversations.

So in order to create chatbot dialogue that is engaging and inviting, we have to appeal to the more evolved part of our brains that is more attuned to the subtleties and nuances of communicating with other people.  One of the best ways to do this is to keep in mind some person that would be the ideal user for your bot.

Don’t think of classes or types of people — think of a real person that you know well.  It would be better to think of your best friend in college, Steve, rather than your bartender.  Then when you are writing, write as if you are talking to Steve.  This way you won’t have to stress about writing for the unwashed masses.  How would you chat with Steve?  Would you send a smiley or other emoji?  Then have your bot say it that way.  Would you tell a short anecdote to illustrate a point?  Do it the same way in your chatbot.  Use jokes to humanize the conversation, if you would talk to your friends the same way.

Sure, chatbots are still a new thing.  As such, we tend to be more interested in getting things to work right behind the scenes.  Too often, the actual written dialogue is more of an afterthought.  This manner of thinking, however, is a grievous mistake.  Think if you opened a restaurant:  There may be some benefit to micromanaging the kitchen, making sure there’s a perfect amount of salt in the entrees.  But you’re not going to waste time with that if your waiters are putting their smokes out on the customer’s plates.  Chatbots are rather similar to restaurants.  Yes, it’s important to serve a great-tasting meal.  But the customer will remember how they were served.  It will last much longer than the taste of the food itself, and it will greatly influence the probability they become repeat customers — or even if they will recommend the experience to their friends.