“Hey Siri, call me an Uber.”
“Hey Alexa, schedule my next haircut.”
“Hey Cortana, play my favorite song.”
Smart AI is no longer a novelty worthy of science fiction; it’s now ingrained into our everyday lives. You could probably recognize Siri or Alexa’s voice in a crowded room, and chances are good that you’ve requested their services on more than one occasion.
As tech giants like Apple, Amazon, and Microsoft continue to enhance their products using AI techniques to keep up with real life demands – integrating with third-party apps like Uber or Spotify, for example – the capabilities of these smart bots seems endless.
In fact, more and more companies are using smart bots and even chatbots to create customized experiences for their shoppers. The ecommerce company Spring, for instance, uses chatbots to connect their website with Facebook Messenger so that customers can browse their inventory directly from Facebook.
But if you’ve ever used any of these voice or chat assistants, you know that the technology is still far from perfect.
Siri, for one, still has problems understanding thick accents. She can perform simple math, but gets hung up on more complicated equations. While she’s great for telling the occasional joke, she still has trouble pinpointing exact locations to make restaurant or shopping suggestions.
But if smart AI is everywhere, why are supposedly “smart” bots still struggling? Why are so many companies still battling to get their bots to do what they want them to do?
The answer: data.
[content_upgrade cu_id=”689″]Don’t Miss: Which Bots Should Your Company Use?[content_upgrade_button]Click Here[/content_upgrade_button][/content_upgrade]
How Bots Learn
The issue isn’t that bots are stupid, exactly, it’s that they don’t always have enough of the right kind of data to learn properly.
Bots are a lot like children: You can’t just give a kid an encyclopedia and expect him to understand everything about how the world works.
First, he has to learn how to read, then he has to learn how to comprehend the things he’s reading, and then he has to learn how to give that information context and apply it to his world. Bots are the same way.
[Tweet “Bots learn the same way people learn – through progression.”]
You can’t just give bots data and expect them to understand what it is, why it’s there, and how to use it. Bots learn the same way people learn – through progression. And, like humans, the more data bots receive, the better the results will be.
AI uses algorithms that take in data and then determines what you want to do with it. The second half of that equation – the “what you want to do with it” – is the context, and every piece of data a bot receives needs to have it.
If you touched a stove, for example, you would learn very quickly that it’s hot. That’s input, or the first half of the equation. The context, in this case, is that you shouldn’t touch the stove again because burning metal on skin causes serious harm.
In order for bots to truly learn, they need the context, not just the input. But that context can’t come from just any type of data.
The Trouble with Data
When people talk about “Big Data”, they generally mean unstructured data – streaming data that comes from social media platforms, mobile applications, location services, or the “Internet of Things”.
Most “unstructured” data sets are actually structured on a micro-level –for example, one data set could show all customers that have purchased a certain product while another data set shows all those who have returned it – problems arise because they may not be structured the same way, making them essentially “unstructured” as a whole.
Unstructured data is useful for many things, like aggregated data for sentiment analysis , multimedia content, sales automation and social media tracking, though the prime example of its usefulness is Google search: unstructured data helps pinpoint keywords in the web pages where they appear.
But the biggest problem with unstructured data is that there’s a lot of it, which can make mining that data for useful information difficult.
If you were going to ask Alexa for articles about Apple the company, for example, but you only asked her to search for “Apple” as a keyword, you might get mixed results between Apple, Inc. and articles about apples, the fruit.
Bots, as smart as they are, can’t always pull the context of your search directly from your brain. If they can’t find the context in the words you use, you have to sift through the search results yourself in order to find what you’re looking for (or use more specific keywords).
But if you were to ask Siri to “Find me shoes under $50”, for example, she would return local search listings with worse results because she does not have direct access to structured data from Amazon. On the other hand, If you were looking for articles about Apple, Inc. on Google, you would probably get results about the company and not the fruit because Google understands the context of those keywords and has access to your search history and other relevant context.
But without access to structured data, neither Alexa nor Google would be able to produce the right results. Sifting through mountains of unstructured data without context to find the right data is like having a child guess the best answers to a complicated question. Sure, they might get it right some of the time, but they won’t learn much from the process.
How to Make Bots Smarter
Structured data makes a bot’s job significantly easier because it’s organized by fixed fields, so search operations or algorithms can find (and interpret) information better and faster. This not only goes for AI assistants like Siri or Alexa, but for any type of internal algorithm in your company.
Because structured data is targeted to specific functions and tasks, bots don’t have to work so hard to understand the process. It essentially takes the guesswork out of it, making the bot smarter in the process.
It can also make businesses smarter by saving time and money, since they can gather useful data quickly and at a much lower expense.
Most businesses simply don’t have the manpower or finances to sort through the proverbial needle in a haystack of unstructured data to find exactly what they’re looking for. And unless you’re a big enough company like Amazon or Google, chances are that you don’t even have access to enough unstructured data to get proper results.
By taking advantage of structured data, on the other hand, you can train your bots to understand the context of your business and your customers more quickly and cost effectively. You’ll know exactly what data to grab and when, and so will your bots.
[content_upgrade cu_id=”689″]Checkout this comparison of bot platforms that your company could use[content_upgrade_button]Learn More[/content_upgrade_button][/content_upgrade]
Final Thoughts
The ultimate goal of any bot is to help you find exactly what you’re looking for and to do exactly what you want to do, whether that’s find an article about gluten-free cooking, playing your favorite music, or ordering up a ride to the airport.
In order to do that, bots will need continued access to a knowledge graph of structured data so they can grow and learn at a steady pace.
For businesses, a bot’s goal is to use structured data to help customers find what they need, when they need it. But more so, by giving customers the chance to find exactly what they need, it can help companies move ahead of their competition.
The original robots who caught my attention had physical human characteristics, or at least a physically visible presence in three dimensions: C3PO and R2D2 form the perfect duo, one modeled to walk and talk like a bookish human, the other with metallic, baby-like cuteness and it’s own language. 
