Many businesses believe Artificial Intelligence (AI) can solve all of their problems or even replace people, yet in 2024 RAND found that 80% of AI endeavors failed to address the problems stakeholders set out to solve. The reality is that, while AI is well suited for some situations, it is likely not the right solution to address all opportunities in your business. So how do you determine if AI is the right solution for the problems you’re looking to solve?
Applying AI as a solution without first clearly defining the problem often results in little return on investment. Understanding the problem clearly and having good data immediately boost the odds of success when implementing AI.
Step 1: What’s your problem?
Every effective solution begins with a deep understanding of the problem. That means clarifying the pain points, knowing who's impacted, and setting measurable goals for success.
We recently worked with a client in the eLearning sector who wanted to leverage AI in curriculum development. Our client had curriculum authors who spent over one month developing curriculum for complex work environments including healthcare and aviation. With ever changing guidelines and the need to quickly and accurately train professionals, the time to market for curriculum development was critical.
What are the pain points?
The time to create training materials and roll these materials out to medical facilities had a significant impact on patient outcomes. Imagine a hospital where new protocols for insertion and maintenance of a central line are not rolled out. The longer it takes to get these new processes rolled out to hospitals, the greater risk to patients.
The ability to quickly disseminate and train medical professionals on advancements in medical procedures and changes in guidelines for care is imperative. Curriculum development was time consuming and required combining information from multiple disparate sources as well as collaborating with subject matter experts and reviewers to ensure the accuracy of the curriculum. This process could take four to six weeks to complete which means that the time from a new guideline for care being approved to medical professionals learning this material can be several months to years.
Who are the people impacted by this problem and the potential solutions?
The process of writing the curriculum was not only long and arduous, but the impact was in the delay of getting information to doctors, nurses and other medical personnel. This was a costly delay since it impacted patient care, treatment and outcomes.
Which KPIs will be used to measure success?
With a better understanding of the process and the pain points, we quantified the problem and established KPIs to help measure success after implementing a solution.
Measuring time spent in curriculum development, time to getting doctors and nurses trained and patient care and outcomes were key indicators to monitor when implementing this solution.
Step 2: Does your data make the grade?
The true power of an AI solution hinges on a critical element that often goes overlooked: the data it's built upon. A cornerstone of our work with our client was assessing the type, quality and availability of data.
AI models used by ChatGPT, Claude, and Gemini are trained on billions of data points that enable behaviors we’ve come to know and expect. However, for specific industry use cases, additional data and training is required so that useful and productive results applicable to each specific use case can be rendered.
Having a dataset is the first step; however, another layer to consider with regard to data is its quality and how well it is maintained. Poor data quality leads to poor AI output and data that is either never updated or uncontrolled leads to outdated or untrustworthy AI results — none of which is ideal when trying to create a highly autonomous solution. The amount of data needed depends on the specific use case, but generally the more relevant data, the better. It is also essential to ensure you have a way to maintain the data you have to guarantee its relevancy and accuracy.
Fortunately for our client, they had a sufficient amount of data on the new guidelines as well as information documented from subject matter experts. The information was accurate, consistent and could easily be summarized. Curriculum creation was based on patterns and conclusions that could be derived from the documentation. The complexity and high volume of the data coupled with the repetitive nature of creating summaries benefited from the speed of AI. Additionally, a procedure and process was in place to ensure updates to the data were implemented and distributed when new changes were published.
Step 3: Is AI the right tool…
With a clear understanding of the problem and the data at hand, we can determine if AI truly holds the key to unlocking a better solution for our clients.
Given the time required to manually create the curriculum and the fact that the data was available for summarizing the information, an AI solution was a perfect fit. With AI, the curriculum development process could be expedited and only require review at the end for quality assurance. Additionally, the process could be more flexible and adaptable to incorporate new research as it becomes available.
Diving deep into the problem space uncovered the true amount of work and investment needed to stand up an AI solution and the supporting infrastructure. Understanding the investment needed combined with KPIs and success metrics, our client was equipped with the tools needed to make an informed decision for the business and has significantly mitigated the risks associated with AI solutions that come from not following these steps.
After consulting with the client to determine if AI was the right solution, we partnered with them to build and launch an AI curriculum tool that reduced curriculum development from four weeks to only nine hours.
What did we learn, exactly?
Through each project, there are a few key takeaways. Understanding the problem clearly and having good data immediately boost the odds of success when implementing AI. In the example above, we learned the key pain point for our client was time to develop and market the new curriculum. We were able to leverage AI to help speed up that process and thus more quickly train medical professionals.
Our experience with this client is just one of many AI integration projects we’ve helped clients bring to life to help teams and businesses go faster and work smarter. From education to healthcare and energy, we’ve helped teams implement AI to accelerate curriculum development, derive insights, and generate content to save teams weeks to months of time thus enabling teams to spend their time and energy more effectively.
It's also important to acknowledge that AI is not for everyone — sometimes it may be cheaper and even more effective using basic automation. Failure will occur and you will need to pivot so it's important that you start small and iterate through continuous development and improvement. Lastly, keep your end goal in mind and maintain a roadmap plan, but expect detours along the way.