Adfast Artificial Intelligence

As part of our collaborative and inclusive innovation, Adfast is proud to announce the integration of artificial intelligence within its operations. This initiative will allow us to improve our production efficiency, reduce lead times, and enhance the customer experience. The use of intelligence within our processes will allow us to predict demand, tailor our product offering, rapidly manufacture and deliver each order, and provide our customers with an experience similar to Amazon’s.

What is Artificial Intelligence?

At first glance, artificial intelligence may seem somewhat cryptic. However, the fundamentals behind these buzz words are simpler to understand than you may think. The following article aims to demystify core concepts such as “A.I.”, “Machine Learning” and “Deep Learning”.


For starters, let’s break down the words “Artificial Intelligence“. All this means is that a computer program can simulate human decisions. A fun example is to imagine a dinner party where an intelligent program selects a recipe based on the dietary restrictions of the guests. After considering the restrictions of each guest, the program will run its algorithm and choose a meal suitable to everyone’s dietary requirements.

At Adfast, our Guided Product Selector is an example of an intelligent program. This tool, which is available on our website, presents a list of questions to our customers about their application, analyzes answers and suggests the Adfast product best suited to their needs. Click here to try our tool!

Now what’s Machine Learning?

At this point, we’re no longer talking about an intelligent program that simply mimics human decision making, but rather a computer that has human learning capabilities. The nuance lies within the task performed by the computer. Instead of asking it to filter through a list of conditional data to make a decision, the computer is presented with a desired result, from which it is asked to create the list of conditional data that will allow it to achieve the result.

The example of a recipe can illustrate this process. In traditional Artificial Intelligence, we would provide the computer with a recipe and ingredients for it to cook a meal. When it comes to Machine Learning, we instead give the robot the prepared dish, then the ingredients, and ask it to figure out the recipe.


How is the computer able to deduce the recipe? It takes the ingredients, combines them together and cooks a dish. It then compares its meal to the initially presented dish that was prepared by a human. In most cases, the robot will not have created the desired result on its first attempt. The process is therefore repeated until the robot’s meal is identical to its human counterpart meal.

In machine learning, a human presence is required to oversee the robot’s learning process by validating if it has indeed learned from its mistakes.  However, this task can quickly become redundant. Researchers have therefore developed a computer self-validation process, which is known as “Deep Learning“.

This final process enables computer to validate its own answers. Each time the computer generates a result, it is compared with a theoretical value. Based on this value, adjustments are made, and the computer increases its understanding of the requested task. This is where neural networks can be applied. The objective is to establish links between the attempts and the desired result. By adjusting these links after each attempt, the computer gradually builds the model (or recipe) that will ultimately allow it to create the desired result.

AI Applications at Adfast

Now that you’ve become an expert in AI, you’re probably wondering “but how can these concepts be applied at Adfast?”
Below are 3 examples:
1- Intelligent production batch scheduling
2- Optimized production equipment utilization
3- Sales predictions

1- Production Batch Scheduling

Production batch scheduling is currently done manually by our master scheduler, Redha. Every day, Redha verifies which products must be scheduled in production according to inventory on hand and customer orders. Production batches are scheduled and associated to different production lines in order to optimize machine occupancy and minimize waste. What waste? Between batches, our packaging lines must be purged to avoid contamination between different product chemistries and colors. If the sequence of colors and chemistries isn’t optimized, a significant percentage of our product is purged rather than sold. During the scheduling process, our planner must consider four important factors: product chemistry, color, format and brand. This practice is time consuming and could be automated.

Thanks to AI, it will be possible to automate the production scheduling process. By analyzing previous production schedules as well as the 4 factors mentioned earlier (product chemistry, color, format, and brand), an intelligent computer program will understand how our master planner makes his decisions and propose schedules that are consistent with our objectives (reduce product waste and set-ups between batches). In addition, it will be able to suggest further improvements.

2- Optimizing the use of production resources

The use of physical and human production resources can also be optimized with AI. An intelligent program can propose a production schedule to maximize efficiency based on each equipment’s production capacity, work hours, available human resources, scheduled corrective maintenance, and scheduled preventive maintenance.

This includes a large amount of information, and therefore our master planner has a very complex task. An intelligent computer program can quickly compute the information, run its algorithm, and propose production schedules that maximizes production efficiency and resource utilization, all while also considering the other 4 factors mentioned above (product chemistry, color, format, and brand).

Unforeseen events, such as equipment breakdown, a shortage of raw materials, or new rush customer orders can happen at any time. Planned production batches must then be quickly readjusted. Our intelligent computer programs will be able to immediately detect changes, run its algorithm, and propose an appropriate solution, unlike a human who naturally requires more time to analyze the situation.

3- Predictive sales forecasting

Once production scheduling and resource utilization are optimized, we can proceed to use AI to predict sales and schedule production batches preemptively. Predicting customer demand will include analyzing customer order history and correlate them to external factors such as weather, the opportunities documented by our sales reps in our CRM, and socio-economic trends. Correlations between sales and different variables is not an obvious task, however an intelligent computer program could increase our ability to predict and react to customer demand. We may, for example, realize that whenever the temperature is above 30 degrees, we sell 25% more 4583 silicone sealant. Or, we may realize that when it is raining, 40% of our customers prefer to have their orders delivered by our Admobile stores on wheels rather than pick-up in one of our Adstores. By analyzing weather forecasts, our intelligent computer program could then preemptively propose that additional 4583 production batches be scheduled or that we double up on stock inside our Admobiles.

When customers place new orders, the desired products will be ready, allowing them to pick-up in one of our Adstores or allowing us to deliver the same day. In the end, the customer experience will be significantly improved, which will set us apart from the competition.