Five Considerations Before Diving Into AI

Artificial intelligence (AI) is all the buzz right now and many businesses are looking for ways to flex their technology and innovation muscles by trying to find a way to implement AI. While AI can provide your business with an edge when applied in the right ways, it’s equally important that you’re not blinded by the “shiny new thing” when trying to find the right solution.

Even as AI technology continues to advance at a rapid pace, it is important for your business to take some things into consideration before diving right in. Here we outline five key considerations when deciding if AI makes sense for your business.

Seeking viable alternatives

New technology spaces, like AI, often bring a lot of excitement with respect to the possibilities, opportunities, and problem-solving potential they possess. That said, it’s important not to get carried away with trying to apply new technology solutions where simpler, more suitable solutions can solve the task at hand. Problems with fewer dimensions or lower complexity levels may be able to be solved without the development of an AI system. Simple yet repetitive tasks, like basic email auto-response systems for example, can often be automated in other ways while complex detection and pattern recognition problems, like facial recognition, are often better suited for AI. Consider a variety of solution architectures, but only act upon those that are correctly scaled and optimized to meet the business’s needs.

Stay focused on the user experience

Reducing human workload by simplifying processes for people is one of the main reasons for developing detection and automation processes. But here’s the key – it must work for the people involved. Human-centered design (HCD) is vital to ensure that the product works for the target audience and end-users. It’s possible to create AI products that work well but do not interface well with the target audiences involved. For example, if you develop a system to reduce the workload of a manufacturing worker but the system fringes upon removing their job role entirely, offers recommendations they don’t feel they can trust, or violates their perceived privacy, they may not be receptive to the product. There is simply no replacement for talking to the people, groups, and stakeholders associated with the problem and understanding their needs in their environments. This will drive whether or not the developed AI system is received well and accomplishes the goals it was developed to achieve. Without using an HCD process, your business may develop and own costly AI products that go unused or even resented by your customers.

Assessing prediction accuracy needs

Many have heard the saying: perfect is the enemy of good enough. In the product development world we call this “minimum viable product.” With AI, that may – or may not – be true depending on your accuracy threshold and your budget constraints. Some environments and problem spaces may be able to tolerate lower prediction accuracy rates. For example, with a weather forecast system, if you’re building a model that predicts the weather with 60% accuracy, it will require far less development time and effort than building a similar model that predicts the weather with 95% accuracy. However, a self-driving car’s pedestrian avoidance system will require greater levels of accuracy, requiring more development time and effort. The higher the margin for error, the lower the cost will be to develop the solution. In general, high complexity high accuracy problem spaces will require higher costs to develop while lower complexity and lower accuracy problem spaces will require lower costs to develop. Today AI is able to extract meaning from multi-dimensional problems with accuracies that were previously not machine achievable. This means that there are likely a whole host of problems AI could solve for your business ranging in complexity and accuracy requirements that are worth considering as your business seeks to innovate.

Leveraging data

Data is a key ingredient in creating AI products. Training deep-learning models and creating meaning from environments requires a lot of data, specifically the right kinds of data for the problem at hand. Your business may have a leg up if it has large amounts of accessible data related to the problem you are trying to solve. If you do not have data readily available to facilitate AI product development, collecting data related to your problem will pave the way toward future potential products and insights. In short, data has enormous power and value, and is the fuel that can lead to viable AI solution and innovative products. Failure to understand and appreciate data as the lynchpin for future success means you may find yourself at odds with the AI development process.

Confirming business experience

Your business’s experience, or industry knowledge is important to consider before developing an AI product. For example, you may operate a manufacturing facility and want an AI solution to predict machine downtime, and perhaps build out an AI solution to provide key insights related to maintenance scheduling. This problem space would align well with your business’s experience and domain knowledge.

If your business experience does not coincide with the types of problems you are trying to solve, you may need to invite others in to help formulate a solution. Building on this example, if the same manufacturing facility wanted to address the effects of financial markets on supplier part costs using AI; then, engaging with a partner who has industry knowledge in financial markets and understands supplier-business relationships would be wise. Overall, if a problem you want to solve already correlates with the types of things your business is already involved in and doing, you will likely have many key insights that could help accelerate the development process. If not, you may want to seek partners who have the necessary knowledge required to solve the problem.

These are starting points to help you determine if AI is a wise investment and what types of problems are best solved with AI. Even if you’re be ready to adopt an AI platform due to lack of infrastructure or team experience, partners (like Twisthink) can come alongside your business to guide you through the process from ideation to system implementation to make your AI solution a reality.