Poorly Defined Goals Put Your Next AI Project at Risk
- Posted by GM, Digital Solutions
- On January 21, 2021
- AI Project, AI Project Goals, Artificial Intelligence, Business Goals, Machine Learning
Any technical project needs clearly defined goals for the best chance at a successful outcome. When using an emerging technology, like artificial intelligence, this issue becomes even more critical. Many AI projects fail to reach their goals due to insufficient planning, which must include a detailed goal-setting process.
Training machine learning models as part of an initial proof of concept helps to eliminate the various unknowns early in a project. Still, trying to complete projects without goals is like taking a road trip without a destination. In the end, how do you define success?
AI projects require a deep understanding of both the underlying business needs and project data. Using a framework such as Machine Learning Canvas can help clearly define the goals and the information needed during the information gathering process. This variation on Lean Canvas provides a template for project communication and project management to help define and reach the team’s desired outcomes.
Since the goals, models, and features are clearly outlined by all teams, the ML Canvas approach helps ensure the necessary buy-in from financial and business stakeholders. It also helps when tracking progress throughout the project timeline.
Here is a detailed look at the importance of well-defined business and technical goals in any AI or machine learning project.
Don’t Let Technical Goals Compete With Business Goals
Technical goals should always be set in relation to the business goals they help to achieve. When there’s a conflict between the two, the business goals must be considered first.
The hype of AI throughout the business world causes companies to focus on including it as a goal of its own. But, its inclusion might not be the best way to meet business goals. For example, placing extra effort to meet a certain technical goal might result in a project lacking sufficient ROI due to the extra costs. Organizations new to the technology sometimes struggle with the business implications of a scaled AI model.
Engaging financial stakeholders early in the project helps keep the focus on ROI as opposed to the latest technical innovations. Ultimately, the technology – even with a nascent innovation like AI – needs to support the business. Convincing the CFO to approve an increase in budget is more likely to happen when business goals are met.
Of course, not all business goals are financial in nature. Environment, social, and governance (ESG) considerations definitely matter when the business goals of an AI project relate to the company’s mission statement. Additionally, some businesses leverage innovative technology projects to raise brand awareness in a competitive marketplace.
Well-Defined Project Goals Also Need to be Realistic
When crafting the technical specifications of an AI project, ensure the business goals are at the forefront. Keep asking the simple question: “does this approach add tangible value to the organization?” Additionally, project goals boast little worth when they are ultimately unrealistic. Because of this, make sure the project doesn’t become too technically challenging; making cost overruns more likely.
This issue becomes more severe at the small-to-medium enterprise (SME), since they don’t have the resources of the large enterprises that successfully implement machine learning projects. While SMEs can still leverage AI and machine learning, the scale and complexity of these projects needs to be a greater consideration.
Using ML Canvas can provide SMEs with a better view of their business goals and the data sources and resources available to them. This allows SMEs to set more realistic goals, guide the development process, and ensure alignment.
Autonomous vehicles provide an interesting example of the importance of setting realistic goals. An SME engaging in a project to build a self-driving car probably doesn’t have the resources to manage the overall complexity of the work, especially including security considerations and legislative issues. On the other hand, an automated tractor project offers fewer safety implications, ultimately becoming a more realistic option for the smaller organization.
Production Deployment Remains the Critical Goal for Most AI Projects
Sometimes stakeholders see a working proof of concept and assume an AI project is then ready for production. In fact, the project could still be a year or more from completion. Keeping goals in mind while considering achievability at scale will ensure your AI project is able to move from POC to production deployment with ease
This same concept also applies to new innovations appearing within the project timeline. When the business or technical goals change, everything must be documented with agreement from all stakeholders. In anticipation of disruptions during the project, create fallback goals and contingency plans to better prepare for these types of situations.
Typically, building a POC is only one intermediate goal on a path to production deployment. Project managers need to ensure all stakeholders fully understand this cold, hard truth. In fact, adding the automation and scalability features to a product likely takes significantly more work than crafting the original concept. It’s another situation where the project goals must be clearly stated. If the end goal only involves building a POC, this needs to be determined upfront.
If your organization has an innovative project idea using AI, a lack of clear goals remains a definite risk. As an experienced AI consultancy, Daitan provides useful advice on completing these types of projects. Our eBook, How to Use Risk Analysis When Evaluating an AI Consulting Partner, puts you on the path to take a project from POC to production.