Not Sure Where to Start with Your AI Strategy? Here Are 3 Steps

Or at least that’s how it feels when it comes to harnessing AI for your business.

The reality is that businesses across industries span the entire spectrum of AI strategy, with some executing fully on several integrated fronts while others are figuratively scratching their heads about what to do with these new technologies, whether GPT-4 or machine-learning-based models.

There’s good reason for all the action around AI. AI investment has risen steadily and is likely to become part of the DNA of most industries. It will soon be table-stakes for staying competitive in most sectors. So it’s critical to get AI strategy right—and soon.

But it’s understandably challenging to know where to start. Or even whether you’ve begun down the right path. To help business leaders think through their nascent or established AI strategy, I’ve broken it down into three key steps, as outlined here. Following these will ensure you’re covering the right bases with AI and setting your business up for success.

Step 1: Know your data (“feed the beast”)

It all starts with data. No data, no effective AI-based strategy or tactics.

As the parenthetical mantra in the section title suggests, you need to feed the beast that is AI, and data is the food of choice. That requires a critical first step of knowing the data your company produces or can access, and what will be most helpful for AI applications.

AI uses a broad range of data: numbers, text, video/images, and others. Your in-house data sources may include email, IMs, Slack, Salesforce, video, audio, sensor data, desk research/surveys, customer calls, and the like. Public data sources, meanwhile, include web, TV, print, consumer databases, digital, and others. So start by gaining a full understanding of the data you have on hand or can get, what would be most helpful to address key questions related to your business strategy, and how to begin storing, managing, caring for, and using the data to get to informative answers.

But here’s the catch: you need data unique to your business; using AI on online data gives you the same generic answers that all your competitors can access. Internal data is the source of competitive advantage when used alone or combined with public data.

Finance companies, for example, have access to huge amounts of public data from government reports, specialized databases such as Compustat, and sources like Bloomberg IMs. But everyone can access those, depending on their budget. Luckily, these businesses can also draw on their own, business-specific data to use on its own or along with public data. It might be internal communication data they can analyze to understand which portfolio managers collaborate best or who on the trading floor is the first to notice a pivotal market change.

Similarly, a hedge fund would have relatively unique portfolio of data that illuminates how much they are embedded in different trading strategies, along with specialized knowledge of lines of business and historical data on how those portfolios have performed. These data can be fed to a machine to find hidden patterns of performance and company-specific capabilities. The AI could be trained to identify market conditions associated with holding a stock too long, buying and selling strategies for particular portfolios, or early warning signs that a trader isn’t right for the firm or may be the firm’s next trading star. Given the power of such analyses, it’s never too early or late to start storing and caring for your data.

Step 2: Get educated to lead (“lead with fire”)

Most everyone agrees AI can make tasks faster, cheaper, and better, whether marketing, manufacturing, or logistics. But the most extraordinary returns come from innovative thinking.

Think fire: Fire solved crucial problems (warmth, protection, illumination) before people understood the chemical properties of fire, because our long-ago ancestors thought creatively about how to use it. AI technicians are trained to implement complicated analyses, but not to know what analyses would be most valuable in the first place. So business leaders must be strategic thinkers who guide AI tech experts, though the leaders themselves may not understand AI’s inner-workings.

What does that mean for you? Commit to getting educated on how AI works (and doesn’t) and what it can do (and can’t) in your business and industry. There are lots of ways to gain knowledge: universities, consultants, industry alliances, and others. Once you do, it will be more straightforward to hire and guide a team of AI experts to manage implementation.

In a large health organization, management instituted an innovative “reverse mentoring” program. Top leaders (including the CEO) were paired with AI technicians (many years their junior) who opened their eyes to the technology’s many applications for classification, prediction, recommendations, sentiment analysis, document discovery, and others. Ultimately the senior executives gained ability to share ideas and language that made discussions of AI strategy more effective and efficient, creating a virtuous cycle driven by situationally aware strategic thinking.

Step 3: Craft your AI strategy (“be thoughtful and comprehensive”)

With Steps 1 and 2 complete or well underway, you can bring it all together into your AI strategy. This usually happens through a combination of in-house discussions, consultants, and critiques of AI applications.

Central to your AI strategy is this question: How can AI strengthen your success and overcome weaknesses? To answer that, understand that AI strategy works on three main types of problems:

• Prediction of success or failure of an action. Here every business might use AI to improve their recruitment strategy by analyzing resumes, cover letters, and letters of recommendation that reveal hidden patterns of employee commitment, ingenuity, and collaborative networks, while reducing unconscious reliance on ascriptive characteristics (age, gender, race, etc.).
  
  
• Classification of objects, people, events, or things to identify resources. Whether your business is high-tech-device manufacturing or farming, AI can be used to classify objects (electronic-component types or avocados about to overripen) or identify the right products to recommend and the right way to recommend them (recommend a new Marvel movie or love story, or recommend the Marvel film but emphasize its love story).
  
  
• Distillation of vast databases for fast answers (GPT-4-style development). Large Language Models (LLMs) are known for their fun, and popular uses. Nevertheless, they can provide serious answers too. Just think how much faster ChatGPT can summarize the key findings of the health benefits of eating walnuts than most people can summarize the same benefits by typing “health benefits of walnuts” into a search engine, choosing links to read and not read, and then integrating the knowledge across links. In medicine, LLMs are being fed huge amounts of research-study data to determine the best next experiment out of a large universe of possible studies. I believe this is the most reliable way to generate the next big medical breakthrough. Shortcuts to knowledge are always valuable, and while LLMs make mistakes, they are capable of continually learning to make fewer mistakes.
  

Use these ideas to get your team thinking how to use AI and the right data to enable creative and innovative directions for your business, while also recognizing the technologies’ limitations, shortcomings, and anticipated trends related to regulation and ethics. The more you can learn about AI and its applications, the better able you’ll be to use these groundbreaking technologies to stimulate profitable growth or diminish risk.

*

This article originally appeared in Forbes.