Bots Grabbed by the Googlies
Google buys API.AI – another nail in the coffin for your customer data
We’ve been saying for a long time that humanlike conversation is the only way to move artificially intelligent applications forward. We’ve also been known to mention on occasion that such a capability isn’t easy to develop either. Apparently Google agrees and have gone out and acquired for themselves something they didn’t have – contextual understanding.
It’s interesting because it sets the bar a little higher for the other tech goliaths to reach and that often means new innovation. It also demonstrates the consumer demand for better, more realistic conversations with technology. But is it going to serve the enterprise need for artificially intelligent applications that deliver meaningful interaction across any platform, service or device?
Techcrunch uses pizza ordering as an example of contextual understanding— “deep dish” and “Chicago-style” can probably mean the same thing to your pizza delivery bot. A little further down in the article they note that according to a counter on API.AI’s website, the company has clocked up over 3 billion conversations. These may seem like two different points, but in actual fact they are not only related to each other, they are dependent.
Simply put there are two main ways your application can know all the different ways to describe pizza bases: code it in (yawn); or machine learn it. Learning is obviously the easiest (I use this term loosely), but you need the data to train the application. Now you see where those three billion conversations come into the equation.
But what happens if you don’t have that data to fall back on?
At Artificial Solutions we take a third option. Our vast language resources are already included within the Teneo platform and provide applications with immediate contextual understanding. Built using sophisticated machine learning, they have been trained based on the billion plus human-machine conversations that have taken place with Teneo-based applications in 35 languages. Our GUI then allows developers to enter two sample questions and point to the right answer. Teneo then works out itself all of the possible parameters. No training data, no linguistic coding, no hassle.
But that’s not all. Integration into back-end systems (yes even the ones you’ve had since before APIs existed) allows enterprises to add to the customer experience by including other relevant data into the conversation. In other words, Teneo not only already knows about pizza bases, it also knows which type your customer prefers.
For global enterprises, foreign language integration is a serious consideration, which is why Teneo is available in 35 languages. Yet it is surprising how few of today’s “intelligent” development applications are available in anything more than the standard European languages. Building new applications for each language is expensive and time consuming which frequently makes projects unviable. The simpler option is to build in the native language of choice and then enable other languages, reusing the existing application with the flexibility to adapt to local culture and laws.
But Google doesn’t just want great conversations to improve training data. Computer Business Review reports—Developers use API.AI’s tools to develop their own machine learning, conversational interfaces for several services. And perhaps here is really the key takeaway from the acquisition. Under new ownership, I’m sure Google will do everything in its power to help developers continue to build their own applications. After all they will run over Google’s systems, potentially giving the company access to your customer data.
I’m not singling Google out here. All of the giants, from Amazon to Facebook have this in mind. They might be reaching for your data in different ways, but their end goal is the same. Understand the customer and deliver what they want. If they can take out the middleman (you) and branch into new revenue streams themselves, so much the better.
Natural language and conversational systems offer businesses a huge potential in reconnecting with their customers on an extremely personalized level. A simple statement – “I’m going on holiday next week” – is an opportunity to sell anything from travel insurance to a book to read on the beach. If you are already using NLI in your conversations then you’ll know exactly what your customer will be looking for and how to sell it to them.
If these conversations are solely allowed to happen over a tech giant’s proprietary service or device, locked into an ecosystem that they control, businesses will lose their most valuable asset—customer data.
I’m not advocating that enterprises should avoid being on Facebook Messenger with some kind of bot. In fact, quite the opposite. I’m saying you need to be wherever your customers are be it Slack, Skype or Oracle to name a few. And not just bots. It’s highly likely in the future customers will want to communicate over their smart home hubs such as Amazon Echo. In order to achieve this, enterprises need to be able to build once and deploy over multiple channels.
Google might be wrapping up this acquisition under the guise of delivering more humanlike conversations but beware, you may be giving away that most important of assets – conversational data that allows you to learn more about your customers.
The way forward is to independently develop enterprise strength artificially intelligent applications with platforms such as Teneo that deliver the flexibility to allow it to be ported to any service or device.
But most importantly, developing independently allows the business to stay in control and maintain ownership of its customer data.