Campofrio Food Group

Artificial intelligence (AI) and machine learning (ML) have become widely available to businesses recently. The availability of these technologies in cloud environments, along with their “API-ification” and integration into ML as a service framework, has allowed businesses of all sizes to benefit from their capabilities.

In addition to being technologically ready regarding availability and accessibility, the point at which technology and business converge is key to these implementations. Without a real business need and improvement, technology has no proper use. That is why the main ingredient in these implementations is to have a team with technical, business, process, and hybrid (integration) profiles that make everything flow.

In the case of Sigma Europe, we have undertaken several initiatives in the field of ML, and it has been a challenging journey. When we started four years ago, ML was seen as a crystal ball, a cocktail to which data was thrown, and it returned what you wanted, typically an improved KPI.

Over time, we have gone through several initiatives in which we have learned to integrate this new technology into our processes and to understand the importance of the path of learning and failure that exists before production deployments, as well as the complexity of finding a good use case that brings real value to the business. Our experience has led us to undertake these projects with a defined roadmap based on three pillars, applicable regardless of the business area associated with the use case.

Methodology

We always start with a pilot project, which aims to know if the production project would benefit the company. It is done in the simplest way in terms of integration but more complex in terms of questioning and discovery.

Given that the phase of problem definition, the variables to be handled, and their correlation, as well as all the discoveries made in exploratory analyses, are significantly changing, pilot projects are developed with agile methodology.

Business questions, as well as requirements, change as we learn from the data. Only when this pilot produces relevant results for the business is a production project proposed.

This is executed in the waterfall model because we already know what we want to develop.

This methodology has helped us to develop more reliable algorithms that are closer to business needs, as well as to have better-defined production deployments faster and with lower development costs.

Data governance and information integration

It is essential to guarantee the quality of the information that feeds the algorithms to avoid erroneous results. That is why all data sources are governed and integrated in a way that guarantees the quality of the data and its exposure to the algorithms. This system can be a data lake or a data warehouse. Part of the data to be governed is also the results generated by the algorithms in terms of deviations to detect potential failure points in the results.

Another key point is integrating information in terms of data flows and systems. These integrations are only addressed in production systems and affect not only the ingestion of data into the algorithms but also the output data. In our case, the information output is taken to transactional systems for operational use and to BI for analysis.

In our case, the benefit of having data governance and integrations is that we have allowed us to have a data flow control system in the generation, consumption, and results, accessible to both IT and business, thanks to which we have visibility of the processes and that allows early alerts in integration or results failures, which allows agile corrective actions. Additionally, thanks to this control, the operational costs of the platform have been reduced to the minimum expression, both for the consumption of cloud resources and for the low volume of incidents.

Change management and culture

Incorporating these new tools into the company has a significant impact on people's work. Fears, reluctance, and resistance can arise, which is why the involvement of business teams most impacted by the change within the project, the communication of the changes and their anticipation, and the explanation of the improvements are essential for adoption by end users.

On the other hand, without an innovative environment, where uniquely doing things is encouraged to create an atmosphere in which people are willing to take risks and fail, everything said above is not viable. New ways of working, addressing problems, and questioning operations and businesses are necessary for these projects to arise in companies.

In the case of Sigma Europe, business and IT teams collaborate side by side in the definitions, implementations, and changes associated with the projects from when they were just an idea throughout the process. We are lucky that continuous improvement and innovation are part of our DNA and are promoted by the company's management.

Let's not lose sight of the fact that the application of all these technologies has enormous potential for businesses. They represent, in many cases, a competitive advantage, and although they begin as a winner, once adopted by all competitors, they will become just a qualifier. That is why it’s so important to bet on them.