Patrycja Welew,

AI in Business Practice: Optimizing Manufacturing Processes vs. Challenges of Trusting AI Tools – a conversation with Greg Bigos, CEO of F33.

Despite the presence of AI in business, especially in marketing and sales, the manufacturing sector is lagging somewhat behind. Could the reasons be problems with access to data, standardization of processes, resistance to investing in modern AI tools, or lack of trust in AI tools? We discussed this today with Greg Bigos, CEO of F33. 
AI in Business Practice: Optimizing Manufacturing Processes vs. Challenges of Trusting AI Tools - a conversation with Greg Bigos, CEO of F33.

In today’s dynamic business world, using tools based on artificial intelligence (AI) is becoming an integral part of companies’ strategies. In the area of production planning, it is becoming a challenge to build trust in modern solutions. Despite the presence of AI in business, especially in marketing and sales, the manufacturing sector is lagging somewhat behind. Could the reasons be problems with access to data, standardization of processes, resistance to investing in modern AI tools, or lack of trust in AI tools? We discussed this today with Greg Bigos, CEO of F33. 

Greg, you have extensive knowledge and experience in the field of production planning. Together with Wit Jakuczun, CTO of F33, you are working on tools based on artificial intelligence with which you want to help companies optimize their business processes. Why this particular direction?

I have been involved in production planning for many years, and our deep experience in this field has led us to choose this particular field. My professional history also includes work on production process optimization, but before that, it was mainly focused on the ERP area, where artificial intelligence was not yet widely used.

Using Wit’s experience in building models and algorithms, we are adding a new dimension to our operations, which is a unique synergy. We have come a long way, innovating on the production side, primarily through the use of artificial intelligence. Continuous improvement and exploration of new possibilities make our approach to production planning modern and effective.

As we mentioned earlier, areas such as sales and marketing are already heavily using AI/ML technologies, and various AI tools have already penetrated many people’s daily lives. However, we note that these technologies are not as widely used in production. Our goal is to change this picture by providing practical and innovative AI-based solutions that deliver actual results for companies operating in the manufacturing field.

Why do you think modern AI-based solutions have not yet convinced manufacturing companies?

There are various reasons why some companies are not using artificial intelligence to its full potential. Often, this is due to simple issues such as not having enough good-quality data to analyze.

Any automation requires standardization. The rules for its operation must be precisely defined so that it can work automatically. Most often, there is one default path called “happy flow,” where automation in production works smoothly. However, when problems arise, manual intervention is required. Implementing standardization can be a challenge, especially in the context of a diverse organization.

Every automation requires standardization, as for something to work automatically, it needs to be defined how it should work. - Greg Bigos, CEO of F33

Organizational flexibility is an asset for small and medium-sized companies. However, the rigidity or standardization of specific business processes, such as warehouse, production, or other operational areas, can often limit a company’s efficiency.

Another challenge is the cost associated with automation. Business owners often find it easier to make decisions to purchase physical tools or automats than to invest in new software. There is still resistance to investing in modern IT tools, and the level of investment in such solutions remains relatively low.

Is one potential challenge to the widespread deployment of artificial intelligence a lack of familiarity with the subject?

The issue of trust in such solutions can also be a problem. When running an efficient and flexible organization with predictable operations and considering replacing human experts with automats, it is necessary to trust that the automat will do the job correctly and fully. Convincing people to trust an automat can be difficult, especially when you pay attention to the other side of the coin – managing expectations. An automat works efficiently, but only as long as processes are standardized, it receives all the necessary information, and that information is accurate. Unlike humans, an automat cannot expertly select information, inquire, question others, and gather information when specific defined data is missing. This is a challenge of trust and managing expectations.

This raises a key question about the cooperation between humans and automats. What are the boundaries between the two areas, and is it possible for automation to completely replace the role of the factory planner?

These limits depend on the specific area. Where everything can be described precisely, automation can successfully replace human labor. On the other hand, in places where intuition is needed, decision-making is based on incomplete data, and in areas of expertise that are difficult to transfer to an automat, automation reaches its limits. Here, automation by itself is not enough.

Currently, a planner working on laying out production plans manually only has time to create one version. He doesn’t have time to develop several variations and compare them. He may be experienced enough to create the perfect plan, but only for one target. He lacks the time to experiment with different variants to see what effects the changes will have. It may turn out that the initially chosen strategy was only seemingly the best.

Artificial Intelligence is a tool like any other. Like a hammer. With a hammer, you can build something useful but also hurt someone or yourself. It depends on your intentions and what you want to use the hammer for. - Greg Bigos, CEO of F33

The “OptiFlow: Factory Planner” optimization tool we are currently launching offers a wide range of capabilities. Its main advantage is the ability to generate multiple production schedules quickly, and each can be optimized for different business goals or strategies. For example, we can eliminate delays in customer deliveries, i.e., on-time order fulfillment, or focus on just-in-time production, minimizing the time products are in stock. Our solution can generate different versions of schedules according to different production scenarios. Moreover, the system can generate these schedules in parallel, allowing them to be compared from different angles. This is an extremely helpful tool for schedulers.

We are working on an “OptiFlow: Factory Planner” production scheduling solution using a chatbot or voice interface. This will enable communication with the application and the calculation engine. If all the data is missing, the application will ask for it, or we can ask for the information we were looking for by looking at various tables or charts.

Previous approaches to automat learning-based solutions have been like black boxes, requiring the right batch of data to generate the expected results. Now, we can more easily obtain the information we need, change assumptions, and adjust algorithms by introducing new variables, targets, or KPIs.

Imagine an S&OP (Sales & Operation Planning) meeting usually held in manufacturing companies once a week. This is a key meeting where decision-makers and representatives from various departments present the current situation in sales and production and discuss optimal solutions. Now, as the head of the company, you can ask our “OptiFlow: Factory Planner” to summarize the situation. You can ask questions like “Tell me more about X?” or “What would happen if Y happened?” the system can answer these questions, providing additional information or generating new business scenarios.

That sounds like a futuristic vision. Can planners already use such solutions?

We want our proprietary optimizer to become a key tool in the production planning process. We can compare it to stacking blocks or assembling a puzzle. Imagine that you know your goal, but the manual task of doing it, i.e., putting the puzzle or blocks together, is the most time-consuming part of the whole process. Our goal is to automate this task. All you have to do is tell our optimizer what you want to get, and it will take care of the rest. Moreover, you can ask him to create several variations of this solution.

Sounds like a big-time saver. Tell us how this could change the role of the planner?

We usually think of jigsaw puzzles as one particular image. However, imagine a set of jigsaw puzzles from which different images can be created. This means that there is no single definitive solution. In our case, the rest of the puzzle is identical except for the edge pieces. We can create various images with a universal pattern that can be customized by rearranging the individual pieces. You can even specify how you want the pieces to be arranged – vertically, horizontally, forming a square, or otherwise.

As a result of this process, the planner’s role takes on a creative dimension. As the decision-maker, you indicate what you want to achieve, leaving the actual putting together of the “puzzle” to our tool. Let’s recall the example with blocks. If you want to build a house out of those blocks, our optimizer will build it for you. Later, you can tell it that the house should have two floors or specify other details, such as the shape of the roof, the number of doors, or the number of chimneys. What’s more, you can provide only a fragment, for example, if you have a chimney and want to add a whole house to it.

We want to steer the role of the planner or process manager here, allowing him or her to set performance assumptions and precisely control certain elements, such as the area of the “house” or its layout.

Could artificial intelligence be wrong, even with access to such an extensive data set?

Of course! Moreover, not only can it be wrong, but it is very susceptible to manipulation. The best example is the saying, “A lie repeated many times becomes the truth.” This is how artificial intelligence works. You can train it to tell untruths. Otherwise – what it finds in its indexed database.

It's not that artificial intelligence will completely replace humans. They will still be needed, just like various means of transportation were used in the past: walking, cycling, and then cars. It's simply matter of making life easier. - Greg Bigos, CEO of F33

Are there effective methods for auditing the results generated by artificial intelligence? Are there special procedures or standards for doing so?

There is no one specifically defined procedure, which is why we are. Our services include defining certain constraints and the space in which artificial intelligence operates. Artificial intelligence is a tool like any other, similar to a hammer. It can be used to build something useful or cause harm. What matters is who will use the tool and how. Many issues related to ethics in artificial intelligence are still poorly defined. Ethics in artificial intelligence concerns awareness of the consequences of using such technologies. That’s why it’s a good idea to use professionals when developing such tools, like learning to drive before driving.

How do you implement your solutions in manufacturing companies?

Our projects start by identifying the business need or the value the solution is expected to bring to the organization. This is crucial. If the tool’s goals are unclear, managing expectations or assessing the project’s success will be more challenging.

During the analysis, it often turns out that the expected solution or problem that was initially defined requires solving a completely different problem. At this stage, we first identify needs and values and then try to understand the business.

I’m talking about projects here generally, not even limited to the “scheduling generator,” which is only part of our field. However, we recently had an interesting example from the food industry, where a manufacturing company had special requirements for how many production lines could simultaneously produce gluten-free food. This was already something unique, specific to that particular company. Despite the standard approach to managing production orders, business presents unique and specific challenges.

Is there a need to customize such solutions for individual organizations during implementation?

Yes, as I mentioned before, automation often requires standardization. Companies that invest in this kind of technology are often successful, indicating their stable market position. Nevertheless, it is important to balance standardization with organizations’ unique characteristics that contribute to their success. If all companies operate on a copy-and-paste basis, they lose their uniqueness. That’s why there is always a need for a customized approach, even if we use publicly available algorithms that sometimes need to be adapted to specific customer needs. The solution needs to change with the changes in a growing organization. Therefore, it is rare to find tools ready to use immediately. More complex business processes require a more personalized approach.

Can implementing artificial intelligence in manufacturing companies change how planners manage production processes and benefit the business?

Yes, introducing artificial intelligence into production scheduling can significantly change the role of planners. Traditionally, a planner could focus on creating a single optimal plan. Still, thanks to artificial intelligence-based tools, he can experiment with different variants of production schedules and quickly evaluate their effectiveness in terms of different business objectives. The scheduler’s role thus takes on a more creative dimension, where he or she can precisely set goals and control certain elements of the process, leaving the actual putting together of the “puzzle” to the newer optimization tool. As a result, companies can achieve greater flexibility, efficiency, and faster response to changing market conditions, directly translating into increased competitiveness and, thus, profits.

Behind F33
Greg Bigos, CEO

Greg Bigos is the CEO of F33, bringing over decades of experience in delivering ERP solutions for manufacturing, logistics, and retail industries.

Contact Greg
Behind F33
Wit Jakuczun, CTO

Wit Jakuczun is the CTO/Chief Data Science Officer at F33 with a PhD in Applied Mathematics and over 18 years of experience in mathematical modeling, data analysis, and simulations.

Contact Wit

Consult us about your needs

Get A tailored 30-minute 1-on-1 consultation with an AI expert.