AI in Your Business: A simple use case with massive returns

Article Originally posted by me on Medium here

The concept of Artificial Intelligence (AI) is nothing new. What is new is its mass adoption across all formats of technology. AI also isn’t always about self driving cars and robots. Google will suggest the best quick response to an email, Flickr will categorize your pictures into cats and dogs based on content, and Tinder bots will automatic swipe potential matches for you waving valuable time when “prospecting”. These are sophisticated examples of AI in consumer apps, though nowhere near the sophistication of self driving cars.

Yet, simple tasks in the business world have been largely untouched by AI practitioners and data scientists. In this post I want to review/reveal some of the most basic examples of low hanging fruit using AI that can have dramatic impacts on your operations.

Customer Service

If you have customers in many various countries chances are you need to provide support in more than one language. How this is typically done is with multiple email addresses (one per language), multiple support pages (localized to each language). Maybe you even have a contact record in a CRM solution where you define the preferred language of a customer. This approach functionally works well, but what happens when you add a customer that speaks a different language than you already offer support in? Well, you start setting up new emails, landing pages, etc. This is not scalable, and puts a lot of burden on the customer to find the appropriate email address to help them. Further, what happens if you start offering support via social media…good luck there…

Why not use AI and/or ML to detect the language that a customer contacts you in? It is extremely easy to develop a language classification algorithm that will accept a block of text and say “this is French”. There are even many (and I mean many…) free services that will do this for you simply by calling an API.

I have personally implemented this solution and it take literally hours to get up and running.

What are the benefits of this?

You are immediately providing value to your customer. The message you send is: “Hey, contact us however you want, in whatever language you want, we’ll take care of the rest”. For your internal team you don’t need to have a triage step where someone physically looks at the message, decides who to route it to, and then does the routing. Automating this process, or digitizingit (#shamelessbuzzwordplug) will save you hours of work, and your employees will love it.

What’s holding you back

With the barrier to entry so low, why aren’t many organizations taking advantage of this? Well, in my experience the concept of ML and AI, though well established in the tech R&D community, has not really made it to the main stream of the business process. That coupled with the complexities of creating ML algorithms from scratch makes the barrier seem high. However, for text classification there are plenty of services that already have this done! All you need to do is call an API.

I won’t list them all here, but services such as Indico.io, IBM Watsoncognitive computing, and HPE Haven OnDemand have ready to use APIs for narrow intelligence driven tasks such as language detection, sentiment analysis, emotional writing tone, image classification, and many more. With these services it is literally a manner of hooking up to them and using your CRM to automate processes with the results (as long as your CRM as capable of it… big if).

For the more advanced use case there are many open source and hosted alternatives where you can build your own algorithm. PredictionIO(recently acquired by Salesforce.com), AzureML, Google Predict, and AWS ML to name a few all provide simple to use interfaces were you can create your own prediction set. For example, use a combination pre-built language detection, text tag/topic analysis, named entity analysis from the likes of Indico.io and your own custom prediction service of case resolution codes you could pre-determine the possible solution to a problem given language, topics in the case, products discussed through named entities, and past solutions from your own custom service. Simply chaining together out of the box services can provide significant improvements in process.

Conclusion

In this article I have literally just scratched the surface of possible use cases of chaining narrow intelligence in customer service. Organizations can seesignificant operational improvements from very simple methods listed above.

In future posts we will continue on this them of discovery of simple use cases in business that can be solved quickly through narrow intelligence. Hope you enjoyed the read!

Disclosure:

I am currently an employee of Salesforce.com. My views do not represent those of my employer and should not be confused as such.