That machines might use algorithms and the power of data analytics to learn, adapt and optimize human decision-making? To an executive looking for a competitive edge — or even just a way to improve business processes — AI sounds almost too good to be true.
The problem, of course, is it is too good to be true, at least for organizations that lack the resources to put AI to work. One recent survey from McKinsey revealed significant annual increases in AI adoption among companies across a wide range of industries, including retail, financial services and the healthcare sector. But this “disruptive” technology, as some have described it, becomes truly disruptive only when certain practices are followed.
For organizations to realize the potential value of AI, according to McKinsey, their AI strategy must align with their corporate strategy. They need a “clear data strategy that supports and enables AI,” and they must invest in staff and technical training to ensure they have the skills to implement AI at scale.
I keep those imperatives in mind when I work with business and IT leaders who wonder how AI might help their organizations. My advice to them: To get the most out of the technology — to make AI work for you — it pays to take a phased approach to implementation.
1. Start with Education
The fuel for AI is Big Data, so the first step involves understanding the various data sources within your organization. Who’s in charge of managing this data? Are there policies and procedures in place around factors such as data governance and privacy? Get a sense of the data challenges your organization faces so you can determine what you’ll need to move forward.
2. Assess Your Systems
Next, take stock of your current collection of data aggregation and processing tools. Do you already have the infrastructure and the expertise you’ll need to launch an AI program? Where does your organization stand when it comes to data management best practices? Are you ready to go, or does your infrastructure need improvement?
3. Move on to Modeling
The AI modeling phase involves developing and training the algorithms you’ll need to address the problems your organization wants to solve. The good news here: There are AI platforms and programming languages you can use that don’t require advanced degrees in data science.
4. Develop Your Infrastructure
Using the findings from your earlier systems assessment, invest in the infrastructure required to quickly turn data into insights. Among the technologies you’re likely to need: storage systems, networking tools, and computing applications and accelerators such as graphics processing units and field-programmable gate arrays.
5. Secure Necessary Services
Few organizations implement an AI program without assistance from an experienced service partner. The choice most organizations face is this: Bring in experts to help with deployment and to teach your team what it needs to succeed, or contract with a partner to handle all elements of the project and guide your AI initiative from beginning to end.
The bottom line is that the revolutionary capabilities made possible with AI are within reach of any organization — as long as you take the time to make sure you do it right.