Many companies today have been swept up in AI fever. They’re not sure why they need AI, they just know they have to have it. And, this is why so many AI projects today are not delivering value − The Wall Street Journal reports that only 53% of AI projects successfully move from proof-of-concept to production.
We recently discussed how a KPMG survey found that many business leaders are not comfortable with how quickly AI is moving and would prefer slowing the pace at which AI is being introduced into their organizations. This, of course, begs the question: how can they break the fever?
The best medicine is to put processes in place for the adoption of AI. Industry regulations and best practices are still in varying stages of development, but this does not preclude companies from adopting basic principles and guidelines for using AI. Here are a few key steps to remember when implementing AI:
- Define Your Needs: Needs assessment is the first stage of any AI implementation but also the most common area where AI projects fail. AI is not a panacea for every technical or business challenge – but it can be extremely powerful for the right challenges.
- Identify Targeted Solutions: Rapid and broad AI adoption runs counter to the most likely route to AI success: to choose a targeted solution for an AI implementation. After conducting a needs assessment, selecting a process – or part of a process – to improve through the use of AI is the best way to have a successful implementation. And, once that targeted solution is deployed, it often helps to identify other areas that could benefit from the use of AI.
- Gain Business Stakeholder Buy-In: A key part of needs assessment is understanding the challenges of various business functions and gaining buy-in from the appropriate business stakeholders on an AI project. For example, if automating loan processing in a bank is identified as an opportunity, getting buy-in from the manager of the loan unit will ensure long-term commitment to the AI application.
- Understand Your Data Resources: AI is all about data. It’s important to understand what types of data are available for the target AI application, considering factors such as: do you own and control the data? Could you collect more, or different data? Is it possible to create synthetic data for the application? AI is like any other process – garbage-in spews garbage-out. So having good data sets, and reviewing and enhancing them over time, is key to AI success.
- Manage Expectations: What is the purpose of the AI project? Is it to generate cost-savings? Or to improve efficiency and automation? It’s important to understand the target business benefits up front and to educate stakeholders in the expected outcomes. If the project is designed to reduce costs, what are the target cost reductions? If it’s to automate a process, what are the benefits of that automation? Setting expectations for projected business benefits will keep everyone on the same page, and make it much easier to report on the relative success of AI projects.
By taking a reasoned, methodical and incremental approach to AI, companies can dramatically improve the odds of having successful implementations …and breaking the fever once and for all!
Download our whitepaper here to explore how we can help you implement AI in your organization today.