AI planning advice for enterprise decision-makers

AI advice

Most enterprises don’t yet have full-blown artificial intelligence (AI) initiatives spanning their business units. Some may be running limited pilot programs, while others still may be assessing how AI can benefit the business before committing valuable enterprise resources.

Indeed, the real-world impact on enterprise strategic goals should be the guiding star of any technology implementation. But there can be many pitfalls along the path to realizing business outcomes through AI. And while there’s a humble nobility to learning from your mistakes, it’s far less costly and time-consuming for enterprises to leverage the hard-won experiences of others by relying on frameworks and best practices to guide them as they integrate AI and machine learning (ML) into processes and operations. To that end, a Harvard Business School article posted this month and a recent study by Lux Research each offer solid advice for decision-makers trying to plan their enterprise’s AI journey – and keep their eye on the goal.

4 rules for avoiding pitfalls

While the Harvard Business School Working Knowledge article, “10 Rules Entrepreneurs Need to Know Before Adopting AI,” is geared toward new businesses, many of these rules make sense for larger and well-established organizations. Among those are:

  • First, understand the business problem you are solving. Your AI implementation must have a definable goal and unique functionality that can improve the business. “Artificial intelligence should enable better solutions,” writes Rocio Wu. AI program results also should be measurable against expectations. Only until an AI initiative has focus and purpose should developers begin building the algorithms that drive the data collection and analysis.
  • Develop your data strategy from day one. AI and ML are impossible without access to high-quality data. Bad data equals bad conclusions and bad decisions. “It’s extremely important to lay out your data strategy from day one—including how things like data sourcing, volume, diversity, privacy and security will be handled,” writes Wu.
  • Manage over- and under-expectations. Although Wu addresses customer expectations, the concept applies to enterprise employees implementing AI and developing/marketing AI-based products or services. There will be employees who are skeptical or even fearful of AI and ML, and there will be others who expect instant miracles. It is important to communicate a realistic vision of both AI’s ability to help them do their jobs and its limitations (particularly in the beginning of the technology’s learning curve).
  • Shift toward a more open and experimental culture. If the goal is to be an AI-first organization, decision-makers must strive to create a data analytics-driven mindset and a culture of experimentation. This allows even larger organizations to be agile and flexible.

4 steps for AI success

In its report on how enterprise can identify challenges and guide successful outcomes for AI initiatives, Lux Research urges organizations to embrace an outcome-focused framework in which they can determine which applications to focus on with today’s tools and how to mitigate challenges preventing successful implementations.”

Lux breaks down AI projects into four major steps:

  • Problem selection – First it’s critical to determine whether your organization even needs AI to solve a specific business challenge. Other technologies may offer simpler and less costly solutions.
  • Data preparation – Automation should be the goal here because cleaning data can be a time-consuming and arduous task. Your data science team’s time and effort are best spent on higher-value activities.
  • Model selection and training – The machine (or model) must be trained properly if it is to make accurate predictions.
  • Deployment – Once your AI initiative is in a real-world environment, organizations must be able to secure data and interpret results.

Step 3, model selection and training, is far more complex than feeding data to an algorithm. The data preparation and model selection/training processes are circular rather than linear. Data determines what algorithms can be chosen for models, while algorithms can influence how data is prepared.

Similarly, deployment (Step 4) at enterprise scale requires fully integrating AI projects into the broader IT infrastructure, including mapping where data comes and determining who owns it, understanding how to detect drift in the data, and deciding how often and when to retrain your AI model. Failure to fully plan AI integration into existing systems has doomed many AI initiatives.

Enjoy the benefits

Enterprises that follow outcomes as their guiding star and take steps to avoid some of the typical planning pitfalls are on course to enjoy the many benefits of AI and data-driven analytics. These enterprises make sure employees understand what AI can realistically deliver – and give them the tools they need to leverage high-quality data and make well-informed decisions.  Then both the AI initiatives and the enterprise can flourish.

Speak Your Mind

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.