man in mid-air, jumping into a body of clear blue water
PHOTO: Jeremy Bishop

More than half of enterprises say artificial intelligence (AI) will play a major role in their organization’s future projects, according to a recent O'Reilly survey. However, the study said, 71 percent of respondents have not yet gotten started. A McKinsey survey found a similar disparity, with companies bullish on AI yet two-thirds reporting they have yet to adopt any AI technologies.

This dichotomy between expectation and implementation isn’t terribly surprising. It takes some time for any transformative technology to move from hype to reality. Everyone didn’t just suddenly jump into cloud computing, for example. The same dynamic is playing out in AI, with early adopters diving in to build new applications in industries like healthcare, automotive, retail and financial services while others merely dip their toes in the water or just wait.

The momentum behind AI initiatives will only grow as company leaders become increasingly aware of the business value that AI can deliver. In fact, the McKinsey study found 45 percent of executives who have yet to invest in AI fear falling behind competitively. As a result, the report said, “companies are twice as likely to embrace AI as they were to adopt new technologies in past technology cycles.”

Many companies that have been standing on the AI sidelines but know they need to get in the game no doubt are asking themselves key questions: What do we need to know about getting started with AI? What are the practical aspects of implementing it in the enterprise? What are common mistakes to avoid?

Here are steps companies can act on to start taking advantage of technology that Gartner predicts will generate $3.9 trillion in business value — in customer experience, new revenue and cost reduction — by 2022.

Related Article: Confused By AI Hype and Fear? You're Not Alone

Starting With AI? Do the Same Things You'd Do With Any New Technology

Pilot a proof-of-concept project to demonstrate AI’s value in a low-risk manner. Typically, the PoC sponsor will be a senior or middle manager on the IT or business side, but it’s crucial to have C-level visibility so the project serves as a showcase for additional and heightened AI investment.

Understand the software development lifecycle (SDLC) for AI applications — planning, systems analysis, systems design, development, testing, implementation and maintenance. As with any application, there needs to be a standardized, efficient and agile method that assures high-quality applications meet defined objectives.

Also understand the product development lifecycle (PDLC) for AI, including requirements, design, manufacturing for hardware and development for software, testing, distribution, use and maintenance, and disposal.

Understand the key technologies and stay on top of the latest trends. AI is a fast-moving area and it’s crucial to maintain deep knowledge.

Related Article: 5 Signs Your Company Isn't Ready for AI

Stay Focused on Your Larger Objectives

AI for AI’s sake makes no sense. Organizations getting started with their first AI projects must keep a sharp focus on the business and/or technical vision and requirements that the new applications are meant to address. Always ask: What problem are we trying to solve? Is it process automation, training datasets to detect fraud, or what? Always start with the end in mind.

Understand that real people make artificial intelligence happen.

Identify the core people who will be responsible for making the organization’s AI vision a reality — the data engineer, the data scientist, the machine learning engineer, the DevOps engineer, etc. More importantly, give those people the tools needed to do their jobs.

Related Article: Democratization of AI Development Is Beginning

Avoid Common AI Mistakes

A number of pitfalls can derail an AI project’s success. These include: not exploring your data enough to know if it’s trustworthy and where there are anomalies; insufficient staff training; not understanding the models used and the underlying probabilities and their impact on the solution; not knowing how to evolve the models to improve underlying probabilities; and using machine learning where it’s simply not needed.

By knowing what to do — and what not to do — companies can begin their AI journeys on solid footing.