Foundations of a successful AI plan

1. Foundations of a successful AI plan

So far, we have discussed how an AI strategist conducts feasibility workshops, ensuring achievable and well-aligned goals through collaboration with stakeholders. Now, it is time to identify what problems AI can solve.

2. AI-suitable problems

Chip Huyen, the author of “Designing Machine Learning Systems”, outlines that AI-suitable systems must be adept at learning complex, evolving patterns. They are best suited for repetitive tasks, providing multiple examples for machines to learn from. Additionally, such systems should be designed to generate predictions at scale, factoring in the considerable investments that go into building such sophisticated systems. As AI may occasionally be incorrect, it’s essential to assess the cost in terms of the potential consequences of wrong predictions and the time to correct them. It is important to ensure that the benefits of correct predictions outweigh the cost of incorrect ones. Benchmarking an AI-powered solution against an existing traditional solution makes perfect sense. Traditional software systems continue providing the existing solution while considering AI to enhance the business KPIs potentially.

3. Assess the supply chain of a furniture manufacturer!

To understand this better, let us consider a furniture manufacturer named OakMingle, where a planner assesses multiple facets of its deep supply chain, such as raw material availability,

4. Assess the supply chain of a furniture manufacturer!

incoming orders, or demand,

5. Assess the supply chain of a furniture manufacturer!

production operations,

6. Assess the supply chain of a furniture manufacturer!

and the end-to-end processing time. These factors determine whether the delivery will be complete or face a shortfall.

7. Any non-AI alternatives?

Before riding on the potential of AI, let's first explore the traditional method to address it. Business rules, such as prioritizing certain customers, might offer non-AI solutions. However, these static rules can not quickly adapt to changing patterns or complex scenarios like when priority is decided based on order value, the number of units, the years of association, or the type of product. AI models, on the other hand, can evolve by learning from past data.

8. Is AI the right fit?

Recognizing the limitations of traditional business rules, we see the potential of AI to solve this problem. Let's assess if AI is the right fit for OakMingle. The model can learn how the planner maintains the demand-supply balance and flags potential shortfalls. With thousands of shortfalls analyzed monthly, there's abundant data for machines to learn. Such shortfalls are repeatedly identified across various products and locations, creating the need for an AI solution to bring efficiencies at scale. Though every mistake has a cost, it is the magnitude of impact that must be accounted for. In this case, the planner can use AI to augment its analysis and identify potential shortfalls to improve efficiency vastly. The benefits outweigh the risks compared to a planner maintaining the entire supply chain.

9. Business to technical mapping

How would we map the business problem to the technical problem? The success of an AI project hinges on the quality of the labeled data. Planners, who are the subject matter experts, label records to train the model, which, in turn, augments the analysis. It's a binary classification problem solvable through machine learning algorithms, indicating potential shortfalls as class 1, else class 0.

10. Economic value assessment

Having established the potential of AI and its technical feasibility, we're now set to make a proposal to the business, which would evaluate whether it is worth funding this project. By leveraging AI, planners can expedite straightforward orders, allowing them to concentrate on complex requests. This would spare them time to follow up with their peers and stakeholders to ensure timely deliveries, streamline operations, and minimize human errors. The result is a tangible ROI shaped by business alignment, feasibility, and potential impact. ROI will be covered in detail in the subsequent lessons.

11. Let's practice!

Businesses face multi-faceted problems. Thus, identifying the right AI opportunities is super critical. With that thought, let's practice.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.