Getting Started With Artificial Intelligence
There are a variety of considerations for organizations as they think about beginning to leverage the capabilities of machine learning and other types of artificial intelligence. The first is to be sure AI is the most effective approach to solving your highest value business problems. That's to say, don't start by deciding to use AI while risking it being a solution looking for a problem or one that doesn't address your most urgent needs. If AI passes this sniff test, the following are several high-level thoughts for beginning this process.
A sensible starting point in exploring the promise of AI is to make an assessment of your organization's data. The quality and breadth of this data will shape the degree to which AI presents significant opportunities. A simple generalization which often holds true is that a company's data will contribute more to the value they can extract from AI than the analytical frameworks they might implement.
The collective data in most organizations is typically housed in a variety of siloed databases. It's often labeled with only regular users in mind, rather than in a way people from other departments can understand and use. And perhaps most importantly, while organizations often only think about their quantitative data, from an AI perspective a company's photos, text files, etc. might offer some of the richest potential value.
In terms of particular applications of AI, a good starting place for organizations as they ramp up their expertise is often with relatively error-tolerant applications. Much of AI is most effective when less than 100% accuracy is required. Sensible expectations regarding this will eliminate much frustration. Examples of first steps with AI which may be easily achievable include applications such as flagging particularly promising sales leads, providing automated chat with non-high value customers, making a first pass at identifying the contents of photos, and evaluating the overall tone in blurbs of text.
Over time, many online businesses in particular will shift from a mobile-first mindset to an AI-first one. But staffing is likely to be incremental and that will usually be the sensible approach. Fortunately companies don't have to invent the entire AI value chain themselves. There is a breadth of AI resources already available, from open data and open-source frameworks to commercial services accessible via APIs offered by numerous companies large and small.
Early hires onto an AI team are likely to include some mix of software engineers and data scientists. But often there will be an equally important contribution to be made from people with product and subject matter expertise. The latter of these can be particularly valuable with things such as thinking through the most important machine learning inputs and planning the layers of a deep learning neural network.