Son, do you use Artificial Intelligence at work? Yes, Mum, of course, that's what we do. Is that the "Chatpete" then, isn't it? No, Mum, not all Artificial Intelligence is ChatGPT. My mother will soon be 70, she's not too concerned about ChatGPT or the latest advances in AI, but she's not the only one who confuses concepts.
In numerous conversations with people holding important positions in relevant companies, we've heard phrases like "we already have a chatbot here" or "everyone uses Copilot or ChatGPT now." Is that it? Does this whole Artificial Intelligence boom boil down to just that? Fortunately, it doesn't all end there.
Forecasting
One of our clients has been using a method for forecasting weekly cement sales for many years, with an average forecast error of 18%. By training an AI model with historical sales data and adding external information such as national holidays or COVID lockdown periods, the forecast error was reduced by more than 8%. It's not as simple as asking ChatGPT or Copilot, but with good work, such interesting results can be achieved. Predicting values is a very powerful tool for any company. In this case, it worked very well for sales, but it can also be applied to product demand, revenue, energy consumption, or staff attendance.
Computer Vision
What company today doesn't have to handle huge amounts of documents? In every opportunity exploration meeting with executive teams, they tell us, "document processing takes us a long time, but we've already automated everything that can be automated." Everything that can be automated? Are you serious? Thanks to RPA, much processing of perfectly structured and defined documents has been automated, but what happens when these documents are not fixed or can change? Good news. They can also be processed. Not everything is possible, but not everything is impossible either. With many clients, we have cross-referenced changing and unstructured documentation. By investing time intelligently and using the appropriate models, incredible results can be obtained.
But it's not all about documentation. Working with images can achieve much more.
-
Material estimation on site: By processing images obtained by drones, we have calculated the weight of material piles at construction sites.
-
Installation verification: By training a model with previous images, the status of installations can be verified and incidents detected.
-
Image classification: Many clients manage hundreds of thousands of very similar images. With image classification models, we have saved them hours of searching and frustration.
Cognitive Agents
"If we were able to implement an insurance policy consultant who could check policy coverages live during a call between an agent and a client, we would reduce client phone time by more than half." No sooner said than done. This statement from a senior executive at an insurance company inspired us to create our agent that processes all types of documents in real-time. It started as an insurance agent, but it can learn anything. Laws, policies, privacy policies, contracts... If it's long and boring, the agent can do it for you. If you want to try one of our agents, you can follow this link:
https://mccminnovations.com/es/eu-ai-act
Optimisation Systems
In the cold winter of Switzerland, after realising that the price of a fondue exceeded €100 and avoiding a bar fight between South Africa and the United States, the most incredible thing that happened to us was hearing "we have a fleet of trucks that doesn't work 100% of the time, and we have subcontracted third-party trucks." It's true that regulations and time restrictions don't make it easy, but that's what technology is for. It can help us process information infinitely faster. Once all conditions and restrictions are obtained, the model generates the optimal delivery route, reducing external contracting by 80% and the fuel consumption of the internal fleet by 20%. Don't you have trucks in your company? This is just one example. Optimisation can be applied to almost any field. Some proven examples:
-
Inventory management
-
Plant production
-
Energy consumption
-
Shift planning
Predictive Maintenance
Most people are reactive. We don't remember the umbrella until we're drenched. We put on sunscreen when we've already gotten burnt. Most companies act the same way. They don't look for solutions until the problem has occurred. Technology can also help us here. A well-trained model can detect failures before they happen. If we fix the fault before it occurs, we get very close to a zero-error culture. It all sounds very theoretical, doesn't it? Aren't there any examples? Many, and real ones. A solar energy production plant detected plate failure when its production decreased. The plate was already failing. A model trained with images before the failure could detect the problem before it happened. The plate was repaired before it failed. Zero errors. In a complex cold generation plant, there are thousands of pipes, valves, and connections that can fail. It's impossible to process all this information manually. But not for AI. By learning how the plant works, it can know what pressure or flow values will cause a failure before it appears.
Do you still believe that AI is just ChatGPT? Book 30 minutes and discover all the possibilities it can open up for you.