Foto: Shutterstock
Foto: Shutterstock

How to define and select valuable AI use cases in manufacturing

Only half of the companies starting an AI pilot project are actually executing it. The key is to choose an idea that will benefit your business. Read more about how!


In 2022, 27% of chief information officers confirmed they deployed artificial intelligence (AI), according to a Gartner AI survey. Even though businesses across all industries are turning to AI and machine learning, prepare your organization before jumping on the AI bandwagon by considering a few factors.

Ask yourself: 

  • Is AI necessary for achieving the project requirements or is there another way?
  • Does your team have the skills to support AI and machine learning?
  • How will AI impact your current operations if you adopt it? 
  • How will you integrate AI with existing systems? 
  • What are the data, security and infrastructure requirements of AI and machine learning?

The Gartner AI survey found that only 54% of projects made it from the pilot phase to production. After significant investment in AI, why aren’t companies deploying it? 

We found the problem begins when companies define a use case. Too often, companies are not identifying AI use cases that benefit their businesses and end-users will adopt. The question is then, how should companies unlock the value and new opportunities AI promises?

It starts with a systematic approach for each stage of the AI life cycle. We developed the Columbus AI Innovation Lab, a comprehensive method to address and account for all challenges when adding AI to your business operations and bring stakeholders into the process at the right time to help you operationalize AI.

The Columbus AI Innovation Lab: Make artificial intelligence work for your business

Define: How to find an AI solution that is right for your business

Successful AI implementation begins with choosing an idea that will benefit your business. Every business has needs and challenges, so an AI implementation that works for another company is not guaranteed to work for yours. So, while a data science professional can brainstorm different ways other businesses use AI, if it is not done systematically with your business goals and needs, chances for success are slim.

Based on our experience helping many customers define ideas for AI uses cases, there are five areas where organizations should focus their efforts to generate ideas for AI implementations.

1. Explore your business goals

One way to uncover and define AI use cases for your organization is to explore your company goals and strategic initiatives. For example, a manufacturer might aim to increase its profit margins by 10% next year. 

It determined three objectives to help them achieve this goal: reduce machine downtime, wastage and supplier irregularities. As part of this initiative, they set preventive maintenance strategies, reducing quality non-conformance products and optimizing suppliers. From that assessment, this manufacturer realized that they could use AI analytics to predict machine failure, product quality failures, optimize supplier routes and rank suppliers. 

These are use cases built from the actual goals of the company.

2. Identify pain points

You can determine AI use cases by looking at the specific pain points your organization experiences. Go directly to your staff to find out what obstacles they face when doing their work.

Identify pain points by: 

  • Conducting workshops with staff
  • Soliciting employee feedback through a survey to find out what the organization can do better

3. Research markets and trends

Another great way to determine valid AI use cases for your organization is to look at your industry’s trends. AI might be your answer if you need to accelerate your solutions to address those trends. Companies can harness the power of data to uncover patterns, anticipate changing buyer behavior and better study market risks with AI.

4. Analyze your competitors

Complete a competitor analysis when looking for ideas to incorporate AI into your business. This helps you understand what your competitors are doing better than your business with various markers, including revenue, profit and more. From that analysis, you can determine if there are opportunities to improve and if AI can benefit you in that goal.

5. Exploring existing data

The data your organization gathers and generates offers insight into where AI could be useful. Many times, historical data can reveal many unrealized facts. When data scientists build connected dashboards across business functions, they can identify gaps when they move from one functional area to another. Then, the data experts work with business leaders to evaluate patterns in the data and do a gap analysis.

For example, we recently worked with one of our manufacturing customers to explore its data. When we did that, the data showed glass breakage happened primarily in the first and last weeks of every month. With the gap identified, we defined analytics use cases.

These five areas will give your organization inspiration and ideas for implementing AI. Smaller organizations might only be able to invest in one use case, so they should focus their evaluation on how AI could help them achieve their business goals. Larger organizations supporting the implementation of many uses might explore each of the five areas to build out relevant use cases for their businesses. 

The Discover phase: Engage key stakeholders in AI decisions

Along with the AI use cases you identified, your data science team may have other ideas from across the company. Before building an AI solution and implementation plan, you must scope and prioritize ideas creating meaningful outcomes for your business. Also important, how can you engage and get the support of the data science team?

To answer these questions and to begin the process of operationalizing AI, business and technical stakeholders must explore the ideas thoroughly, being certain that they support the business KPIs. Without that clarity, a team is far too likely to judge an AI solution’s success by the model’s performance rather than by its impact on the business. In the Columbus AI Innovation Lab framework, this four-dimension evaluation involves all stakeholders in decision-making.

1. Is the AI project contributing to reaching business goals?

When evaluating an AI use case’s relevancy, business stakeholders, such as the executive team, subject matter experts, finance and more, need to determine if the use case will contribute to the organization’s desired future state. This perspective also looks for quantifiable high-level business outcomes for the use case.

In our previous example, where the goal was to increase margins by 10%, the business viewpoint would assess if implementing an AI solution would help the company achieve this goal. 

2. Is the AI project viable?

Next, evaluate the tasks and activities required to realize the capabilities identified from the business viewpoint. So, even if the idea is very good and would increase margins by 10% it’s not a good use case to pursue if it doesn’t pass the usage evaluation. However, if there is a way a company can use AI to achieve the objective and the use case is deemed viable by stakeholders, then it’s a good use case.

For example, demand forecasting with AI is a method this company believes is a way to increase margins by 10% by identifying the demand and then offering a discount to push sales past the demand. However, the dealers said they couldn’t increase demand regardless of the extra discount. This use case would fail the usage evaluation. Even a good idea to use AI might not yield results for the company if it is not useful.

3. Do you have the organizational resources?

With the functional assessment, the team evaluates a system structure, interrelations, interfaces and interactions between components and external systems. From the functional viewpoint, there are:

  • Five functional domains (industrial control, operations, information, application and business)
  • Five system characteristics (safety, security, resilience, reliability, privacy and scalability)
  • Four cross-cutting functions (connectivity, distributed data management, industrial analytics, and intelligent and resilient control)

For example, if you want to meet the growing demand identified by demand forecasting by increasing production, but you can’t add shifts because your production team is already working three shifts. Therefore, the only way to increase production is a much larger investment, such as building a new factory. As a result, this use case would not be helpful to the organization’s initial goal of increasing margins by 10%.

4. Implementation: Do you have the right AI capabilities?

Even if you have a great idea to use AI in your organization and a huge data set, if you do not have the technologies, talent, systems and procedures to make it happen, it is also not a feasible AI use case for your organization.

For example, your proposed AI use case requires high-performance computers, and your organization currently does not have the data to support implementation. This would make even a good use case fail from the implementation viewpoint. 

Evaluate effort and value

Next, you will need to prioritize each use case by evaluating the effort it takes to build and adopt the solutions and the value the solutions bring to your organization. In our model, the matrix will have four quadrants:

  • Easy wins = Less effort and more value
  • Big bets = More effort and more value
  • Maybes = Less effort and less value
  • Thankless tasks = More effort and less value

Obviously, organizations would want to immediately implement the use cases that fall into the easy win quadrant and de-prioritize those in the thankless tasks quadrant.

After prioritizing use cases, it is time to create the road map outlining the people, process, data and technology required and the timing for each phase. We will present that in our next blogs.

When you follow the Columbus AI Innovation Lab framework and complete the Define and Discover steps, you have a solid foundation for a successful AI project.