Artificial intelligence · Marketing

What's your take on the recent wave of AI and Chatbots in every industry?

Hira Saeed Digital Marketing Strategist | Blogging | Event Management | Communication Specialist | PR | Public Speaking

November 2nd, 2016

I have been exploring Artificial Intelligence and chatbots industry for quite a time and loved the level of innovation it is pouring out. Being a marketer, I see things differently than from the perspective of a developer. I want to ask from the fellow members here, what's your take on this wave? How canwe cash out this industry to make their products and services better?

Joe Dew Product Management - eCommerce, marketplaces, and machine learning

November 2nd, 2016

Hi Hira-

I've been working in the deep learning / AI space for several years. While the overall promise of AI is compelling, its practical application for most start-ups today is still pretty limited. Here are some considerations: 

  • Data: a big constraint is access to data. Yes, there are generic libraries out there than can help you get started on building a really rudimentary machine learning app. But the true power of AI is the ability for the machine to learn through constant training. This process requires A LOT of data. That's why most of the players in this space are the Google, Baidu, Apple, and Facebooks of the world who have access to tons of data. And it's not just raw data. You need access to correctly labeled data to do your benchmarking and testing. I think there could be an opportunity to build very narrow AI applications that use a very deep but constrained set of data. (e.g. healthcare, self-driving cars, internal enterprise data, location data, translations, online advertising, etc.)
  • Precision. There's a big difference between 90% precision and 98% precision, meaning the accuracy of a result. We've all experienced Siri, which is generally pretty good but not good enough to work consistently. This deficit often means users stop asking long tail queries. So again, if your use case is sufficiently constrained (e.g. show me every image with a cat in it), you can make it work. But having a general purpose chatbot that works consistently is really difficult to build.
  • Domain expertise. Assuming you can access data and have a working production system, you need to employ some machine learning experts who know how to tune the algorithms to fit your use case. AI system are often non deterministic, meaning that it's not always obvious why the system responded in the way it did. To understand requires a lot of tinkering and iterating. AI is not a magic black box that just works.
  • Business model. The jury is still out on the economic viability for chatbots. I suppose if it replaces human call center workers it could work. In general, AI is good for three things: 1) answering questions that would take a human <2 seconds to do (e.g. what's in this picture?); 2) predicting behavior based on a sequence of data (e.g. clicks on online advertising, investing, polling); 3) recognizing patterns in a large set of data that no human would ever figure out. (e.g. fraud detection, spam). There's probably money to be made somewhere in there.
Hope this helped. If there's a specific use case you're thinking about, I'd be more than happy to help.

Andy Freeman Product Management and ... - Looking for new opportunities

November 3rd, 2016

I think that most people are looking at chatbots incorrectly.  AI isn't the important aspect.

Remember the web?  You didn't "install" yahoo - you visited it.

Apps are different - you install them.

It's no big deal to visit my car insurance company's web-site twice a year to pay my bill.  However, it is absurd to install an app to do that.

Chatbots, like the web, are a low-commitment interaction model.  Apps are a high-commitment interaction model.

Many relationships are better served with a low-commitment model.

Chatbots do tend to be more conversational than either the web or apps and AI can be helpful in a conversational interface.  However, one can do a good conversational interface without AI.

Amr Selim Customer Success Expert | Technology Consultant | Keynote Speaker. I help businesses grow by improving CX & Performance.

November 2nd, 2016

I keep seeing this question, which tells me a lot of people are cautious about this and they should be.

By the way the AI or Chat bots isn't a new idea, it just got refined and re-branded, some forms of it existed in many help desks of the past as part of their self services. Call centers have had voice activated IVR (Interactive voice response) for sometime now.

It all depends on the end goal of implementing it, fundamentally, why would a business jump into that bandwagon ?

Unfortunately I have witnessed some of the technology vendors citing "cost savings" as this technology would result in the ability to serve more customers with less "Human Resources" ! Such rhetoric is very dangerous and will definitely lead to a high churn of customers.

The goal isn't supposed to be discouraging or delaying a customer interaction with a service or sales employee, it is supposed to aid that interaction and provide both parties with important information that will enhance that interaction. 

Very much like the CRM rush of the 90's, everyone jumped on the bandwagon then complained that it did not work for them! 
Of course it did not, because you can't automate a process that doesn't exist !

So first, you need to go to the drawing board and jot down the final goals + all possible scenarios / all channels of interactions, then come up with the right process that should lead to achieving this goal.
Test it internally then tested as a pilot with a few selected customers, once you are sure that it all works well, automate it using the technology or tool of choice.

Amr Selim Customer Success Expert | Technology Consultant | Keynote Speaker. I help businesses grow by improving CX & Performance.

November 3rd, 2016

Spot-on Andy,
90% of customer experience comes from the Human Interaction! Technology has a great role in facilitating and enhancing that interaction, not try to replace it.