Executives must understand AI before their company can be AI capable

GUEST: Artificial intelligence has become the buzzword du jour.  At Google’s I/O 2017, the company announced its slogan change from “Mobile First” to “AI First,” fueling the frenzy. VCs, startups, and established companies alike are all getting into the gold rush of AI. But what is AI, anyway? First, what AI is not: AI is not one […]

---

GUEST:

Artificial intelligence has become the buzzword du jour.  At Google’s I/O 2017, the company announced its slogan change from “Mobile First” to “AI First,” fueling the frenzy. VCs, startups, and established companies alike are all getting into the gold rush of AI. But what is AI, anyway?

First, what AI is not: AI is not one single technology.  AI is made up of multiple technologies. The technologies under the AI umbrella are at varying stages of maturity — and promise — when we think about broader business application. The two getting the most attention at the moment are machine learning and deep learning. Think of these as smaller subsets of each other, with artificial intelligence being the umbrella, machine learning being a subset of that, and deep learning again a subset of machine learning. Advancements in deep learning (which employs neural networks to “learn”) the past few years has sparked the most excitement in the industry and also shows the most promise.

Second, AI is not a single industry, but rather a layered industry comprising a data layer, a statistical layer where models are built, and a business application layer. When a startup says it is an “AI company,” it’s helpful to understand which layer they operate in because one layer alone may not give you what you need.

So, what is AI?

At the heart of it, applied AI is about patterns and predictions. It’s how computational machines can recognize images, texts, speech — essentially large sets of structured or unstructured data — to determine patterns and consequently make predictions. What is the likelihood this customer will purchase a car after buying a new home? Could this MRI image show malignant cancer? Is this credit card charge a fraudulent transaction? These are some of the types of predictions borne out of AI technological applications.

The underlying technologies include natural language processing, image recognition, speech recognition, computer vision, and much more. The core of these technologies is the ability of the computer to “understand” something, whether text, speech, or image, and deliver a task or output.

AI has been around for more than 50 years, but the advent of stronger computing power and the availability of vast amounts of data has enabled a breakthrough in artificially intelligent offerings. From Deep Mind’s AlphaGo beating legendary Go player Ke Jie to Amazon’s Alexa advancements recently, the developments in the past three years have been revolutionary.

Digital transformation bears fruit

Companies want to cash in on the gold rush, but what are the building blocks that allow a company to be AI-ready or, at the very least, AI-capable? Any non-digitally native company has several steps to take before capitalizing on the next wave of artificial intelligence application, the most important being its digital transformation.

When a company embarks on a digital transformation, it migrates from an analog world to a digital one, where transactions and information are digitized. A full end-to-end digital transformation also includes upgraded technological capabilities, such as cloud migration and re-imagined processes that allow better agility and access. The digitized data becomes an asset that allows models to be built that then predict a desired outcome. The vaster the amount of digitized data a company has, the better its ability to predict an outcome. The more technological capabilities a company has, the shorter the cycle times, which enables faster learning.

This can be illustrated further by the virtuous cycle of data enabling product: Having users or customers that provide data allows that data to inform better product development. Better product development means a better product that then attracts more users. And so the cycle continues, and the snowball grows.

Successful AI deployments

What are some use cases of businesses deploying AI today? Some of the more popular ones I’ve seen are the deployment of chatbot technology. Capital One, a leading financial service provider in the U.S., rolled out Eno (“One” spelled in reverse), a bot that is NLP capable, allowing the use of SMS as an interface.

Capital One also launched account access through Amazon’s Alexa platform, allowing users to perform a variety of tasks like bill pay, balanced review, and a host of other voice-based transactions.

Stripe, a unicorn in the payments space, launched Radar, its anti-fraud product that uses machine learning to detect aberrant behavior. Machine learning is particularly useful in risk management, as well as fraud.

The examples of AI application across verticals and functions are too numerous to list here, but this gives a flavor of what is possible.

Steps to effective AI integration

So, in the hopes of staying on the frontier, what can a company do? What are the steps we need to take to be ready for the next curve?

1. First things first, review your digital transformation roadmap. How has your business model shifted to align towards the customer, and how have your systems aligned towards data as an asset? Digital transformation is truly foundational to any type of AI adoption in the future.
2. Next, educate yourself and your executive leadership team. How well do you understand AI?
3. Develop your AI strategy. What’s your endgame? In-house or outsourced? Watson, Salesforce, startups? This can take on so many forms.
4. Review your organizational readiness and team capabilities. Do you have the right talent, or are you just shifting resources within your team? Are you keeping your engineers trained on the latest techniques and tools? Do you have a Python developer, NLP statistician, and so on? Is strategic data acquisition a need? Do you have a central data warehouse?
5. View your end-to-end process. How ready are you to develop and deploy AI?
6. Discuss legal and ethical issues at the beginning. While I believe fears of superhuman intelligence are unfounded today, there should be guidelines to ensure the transparency of model development and any automated processes that can adversely impact the customer.

Telegram is an instant messenger that focuses on speed and security. But could Telegram security be compromised with a Telegram messenger spy? Many people would want to Telegram hack to get specific information about their kids, partner, or employees.

How do you see AI playing a role in your company’s digital transformation?