Decision Process

Decision-Makers Approach to Artificial Intelligence

This section assists organizational decision-makers in assessing the suitability of AI technology to their specific requirements while providing business ROI. As discussed earlier, there are numerous AI tools that can be applied to a wide range of questions. There is no AI Swiss Army knife solution. High-level AI applications are easy to imagine, are enjoying a lot of hype, and can be costly if they succeed and even more costly if they fail. Building a business case for AI for your specific requirements is the only viable approach to weed out the good, bad, or unnecessary AI application. This section assists someone considering a new AI application in assessing its suitability for a specific requirement. The following sections are intended to aid this process.

Does the application automate a specific task or activity?

Defining the specific tasks or activities to be improved focuses the application's requirement. While AI technologies can aid a wide variety of complex problems, the first step in evaluating their benefits is to specify what will be expected.

Does the application involve inputs from available digital source information?

The AI application will need electronic source information to process. This information can be from existing data repositories or from new data that can be collected from sensors.

Are there definable outcomes from the application?

The next step is to identify the specific results to be produced by the application. These can include a combination of situation-dependent or other complex outputs, but they must be definable.

Is there describable business logic to be implemented in the application?

The business logic to implement the application will vary depending on the nature of the problem, but the user must be able to describe how the results are to be generated from the available inputs. As described earlier in this topic, different types of AI technologies approach solutions in different ways, but they all require domain knowledge to automate.

Can an existing non-AI tool or product implement the business logic?

Not all problems require a dynamic AI solution. Simple solutions may be possible with static automated logic.

Can the resulting solution be integrated into the intended users' operational environment?

To be operationally suitable for deployment in a production environment, the AI application must be suitable for the intended user's environment, including considerations of physical and logical integration. For example, integration could be a software app or embedded in an electronic product. This integration should also include user-suitability including training or skill levels required to use the application, and any potential impediments to adoption: conceptual, physical, regulatory, or cultural.

Confident that you are ready to activate an AI application?

Still undecided? Consider a pilot project.