Five Steps to Take to Give Your AI Project a Fighting Chance at Success

About the Author:

Kellye Whitney
Kellye Whitney, is an award-winning writer and editor. The former editor for Chief Learning Officer magazine is now the founder and Chief Creative Officer for Kellye Media, a Chicago-based media coaching, content and consulting company.

Five Steps to Take to Give Your AI Project a Fighting Chance at Success

In 2018, industry research firm Gartner made a rather disheartening prediction: Some 85 percent of AI projects are doomed to failure. Wow. That’s a lot.

Keep that number at the back of your mind. Now consider how many questions an organization must answer when it comes to this kind of project implementation: Do we have enough data? If not, does that data exist, and how will we gather and categorize it? Are the algorithms powering our systems sound? At a foundational level, what makes more sense now that we’re considering implementing AI? Should we adopt certain features from existing software platforms, or build a customized model?

The answer to that last question is, it depends. The decision to replace known legacy systems and processes with a new AI solution is not an easy or a quick one. Customization is always a good idea when it comes to maximizing results for a specific organization, but there are obvious drawbacks in both the associated risk and the expenses related to appropriate staffing, development, testing and deployment.

After all, investments in the time and training needed to develop successful AI projects don’t come without cost. Further, when you start from scratch, there are no guarantees because there is no previous foundation of success from which to borrow.

To increase the likelihood of AI project success, organizations can do the following five things:

1.   Lay the right foundation: Assessment is key before you begin an AI project. Employees and customers already know how products and processes work, which means they have expectations about what you can deliver. This makes the decision to take a chance and change a legacy system in favor of an AI implementation even more difficult.

This initial assessment phase takes time and can be costly, but it’s critical that leaders realistically assess what exactly they have to build in order to lay the foundation for project success. The results of that assessment will come in handy. According to a study from Pactera Technologies, a global technology company, some 77% of top tech leaders indicate that pushback from senior management, and their failure to impress CIOs, often stand in the way of their efforts to get new AI projects up and running.

2.  Determine what’s right for the business: To mitigate some of the risks associated with a new AI project, and reduce the chance of failure, ask – and answer – key questions:

  • Why do we want to use AI?
  • Are we trying to build new, better products?
  • Are we trying to increase our time to market?
  • Are we trying to use AI to increase efficiency?
  • Are we trying to use AI to decrease security or compliance risks?
  • Are we trying to use AI to increase profitability in specific areas of the business?
  • Would another type of project produce a similar, positive result?

The answers to these questions will help to drive a focused AI project strategy.

3.  Measurement matters: At the end of the day, AI is another technical tool the business has to achieve critical milestones or objectives. So, as is the case with any new project, measurement is important. Leaders have to establish early what success looks like, how they will track the AI project or experiment, and at which intervals, and then have a transparent method in place to audit and report out developments regularly.

There’s an old adage that says, what gets measured matters. Tech leaders don’t want a lack of measurement to make it easy for their peers to abandon the AI project at the first bump in the road.

4.  Train and collaborate: AI tools are evolving rapidly, but to use them effectively, technical professionals need data science skills. Because this technology is still evolving, however, it’s tough to hire individuals with the depth of skill that companies require. So, the logical solution is to train existing talent.

If your organization doesn’t have the skills needed in house, look for a training company that has deep expertise in AI and real-world experienced instructors to offer. It’s also important that the aforementioned company act as a business partner, so collaboration is critical. Technical training partners like Colorado-based DevelopIntelligence, for example, can facilitate an organization’s change management efforts as well as create a customized curriculum that will provide the necessary AI skills technical talent need to meet specific business goals.

Training alone is not enough. Your training partner should be willing to become intimately familiar with your organization’s business strategy, and align any training offered to the specific individuals who take it. That way, the people in those key roles will be able to advance the organization’s business strategy and goals more quickly post-training.

5.  Celebrate the win: Because AI projects are often fraught with difficulty, when they succeed, it’s definitely cause for celebration. So, when projects work, make sure senior leaders know it in a timely fashion. Keeping stakeholders abreast of developments will help to increase and sustain their support and engagement for the next project iteration.

AI projects are not intended to provide quick solutions. They’re best suited to enhance and help build a long-term business strategy. To increase the chances that your AI project will be a success – and keep that dismal 85 percent failure specter at bay – keep these five steps in mind. Communicate consistently, set expectations early and realistically, establish the business case, and outsource to find the training and support your organization may lack. Good luck!