motorcyclist in the air
PHOTO: Darren Nunis

Artificial Intelligence (AI) and its close cousins machine learning (ML) and deep learning (DL) are already everywhere in our daily lives. The technologies power everything from gaming the commute with Waze to binge-watching with Netflix. And in the financial markets, tech players such as Nvidia see their stock soaring due to AI processing demand. Indeed, amongst analysts Gartner, Forrester and IDC, AI dominates 2018 trend predictions.

AI isn’t really “new” technology, but the connectivity, data access and cloud power of our age has finally made it practical and consumable. Yet according to a recent Gartner report (registration required), only 20 percent of CIOs report experimenting with AI today. And only 20 percent of those — just 4 percent of organizations overall — report successful AI deployments.

This means that despite its rising prevalence, 80 percent of organizations have not yet commenced their AI journey. 

With AI hype so high, why is enterprise adoption not moving apace?

For many organizations, the challenge is finding the right formula to jumpstart AI success. The following steps outline the process of preparation, planning, innovation and education required to jumpstart AI in the enterprise.

1. Identify Competitive Advantage

With every major new technology wave — from desktop computing to the internet to big data to mobile — the proverbial technology cart inevitably gets ahead of the business value horse. But business value, typically in the form of disruptive new products, new markets and new cost models, eventually regains its primacy. AI is no exception. 

In its report, Gartner advised CIOs to treat AI as they would any other advanced technology: Start with a business and IT collaboration for pinpointing the scenarios required to create competitive advantage, and then use these scenarios to inform where you might best apply AI.

2. Select Specific AI Application

The next step is to identify the specific AI applications that will deliver your prioritized outcomes. This won’t be easy, given that McKinsey recently identified nearly 600 distinct AI applications spanning the value chains of many major industries. A few examples include planning and forecasting, assembly and manufacturing, promotion and personalized user experience.

AI is not a business panacea. Your objective is to shortlist five to 10 AI application candidates, then select two or three of those as pilots. Based on pilot efforts, move forward with one AI application that can deliver both the business outcomes you seek today and the extensibility to serve the people, process and technology you will need for the future.

3. Procure AI Technology

AI applications, by definition, are built using AI technology. Finding the right AI technology to power your AI application is a critical next step.

Given the rapidly evolving nature of the technology, it is far too easy to make costly procurement mistakes. Fortunately, there are two shortcuts that improve the odds of success.

First, you can take advantage of powerful cloud-based platforms that provide modern AI services running on AI-optimized hardware. Amazon Web Services (AWS) Machine LearningGoogle Cloud AI and Microsoft Azure Machine Learning are three of the most popular and complete platforms. You can affordably use these to quickly obtain a rich set of AI capabilities in support of a wide range of AI applications.

Further, you don’t need to build your own AI applications. Instead, you can procure software with “AI inside.” Today’s business-intelligence and analytics applications embed AI to automatically discover data relationships, identify patterns, and recommend analyses and visualizations.

4. Develop AI Skills

Whether you decide to build your own or leverage an “AI inside” application, your staff will need to acquire new skills. Invest in developing your early adopters, but don’t worry about turning them into AI experts overnight.

To build foundational knowledge, take advantage of open online course providers. Popular resources include:

All of the cloud-based AI platforms also include enablement tools to simplify and accelerate AI application development. AWS SageMaker lets data scientists and developers quickly and easily build, train and deploy machine learning models with high-performance algorithms, broad framework support, and one-click training. Google provides videovision and speech recognition APIs that use powerful neural network models for easily searching and extracting metadata from videos, understanding content from an image, or converting audio to text. Microsoft Azure Machine Learning Studio is a fully managed cloud service that enables you to build, deploy, and share predictive analytics solutions.

“AI inside” applications also provide a range of tools to speed solution readiness. Off-the-shelf solution accelerators include pre-built models and business rules, best practice guidance, and more. Reusable and extensible, such packages allow you to build on the shoulders of others to quickly ramp up your AI applications.

With AI technology, education, and enablement tools in hand, you’ll be ready to implement your initial AI application. But as with any area of innovation, do not expect that everything will go perfectly from the start. When inevitable problems arise, recall the wisdom of Thomas A. Edison: “I have not failed. I've just found 10,000 ways that won't work.”  

Luckily, there’s already community support available. The Artificial Intelligence Forum is a popular site for discussions that range from simple AI overviews to detailed algorithm dissections and specific AI problem sets. Your developers will also find GitHub a valuable source. With over 10,000 searchable AI projects, your team can potentially learn from and collaborate with millions of other developers.

5. Build on Success

In the words of Antoine de Saint-Exupéry, “True happiness comes from the joy of deeds well done, the zest of creating things new.” So as soon as you complete your first AI application, it is time to celebrate.

And because success breeds success, it’s important to market your solution beyond the initial AI team. Write an internal case study that identifies the business problem and technical challenge. Describe the solution, with a focus on how you used AI to execute. External promotion is valuable as well — call on one of your AI vendors’ marketing teams and jointly showcase your success.

Once you have completed more applications, you can begin to think broadly about overall AI strategy and execution. This is also the time when you can expand your team beyond the early adopters, leveraging their experience as player-coaches. And while you’re at it, take a look at lessons learned and best practices. These will be a great foundation for the next chapter of your AI playbook.

As with any emerging technology, AI use will continue to grow and evolve over the next few years. However, it’s already evident that the technology is transformative and it’s here to stay. If organizations can find the right formula for AI-driven innovation now, there’s no limit to what they will be able to achieve in the future.