Monica Rogati a écrit à propos de l’ article “The AI Hierarchy of Needs” sur Medium il y a 45 mois et a partagé cette pyramide perspicace demandant “Comment dire aux entreprises qu’elles ne sont pas prêtes pour l’IA sans paraître (ou être) élitiste – un gardien autoproclamé ?

Truth being said, a lot of companies still do not realize that it is “data” which is at the core of systems such as Data Sciences and Artificial Intelligence. Many organizations continue to fail when they attempt a major analytics initiative. The reason is that the focus tends to be on the top 3 layers of the pyramid above and not on the rest, which lay the foundation to building a strong “data analytics” strategy.

While a lot of organizations are paying attention to building data models, they tend to ignore things such as data quality and data integrity. “Data will redefine how we think about models,” says Fei-Fei Li, Chief Scientist at Google Cloud, and a Professor at Stanford.

Organizations need to ensure that they have good data management processes. “The best algorithm wouldn’t work well if the data it learned from didn’t reflect the real world.” A better dataset is the first requirement before starting off full steam on Data Visualization, Data Sciences, Machine Learning, or Artificial Intelligence.

While most businesses have a desire for data-driven insight, many are not realizing that ambition. The result is that data management is often fragmented and driven by multiple stakeholders. This leaves organizations dealing with a high degree of inaccurate and disparate data and there are several challenges to maintaining it.

So, what should organizations take into consideration when building a data analytics roadmap or strategy?

Volume de données

Nous savons tous que le « Data Analytics » nécessite de « bonnes données », mais avant cela, nous avons besoin de volume. Beaucoup d’organisations ne semblent pas prendre au sérieux le problème du volume. Ils pensent ou supposent en quelque sorte qu’il est disponible. Comme le dit Jana, architecte technologique en chef de Zuci, “Construire une banque de données est un problème d’entreprise aujourd’hui et je pense que très peu d’organisations le prennent au sérieux, même si la transformation numérique est prise avec un sérieux absolu.”

Les organisations ont besoin des bonnes personnes qui disposent de temps au début des projets pour créer des systèmes robustes et efficaces de collecte, de conservation et de stockage des données. C’est l’objectif commercial le plus important de l’équipe des données, pas seulement la construction d’un entrepôt de données et tout y déposer.

Data Quality

The follow-up issue to data volume is data quality. Poor data quality leads to the inability to develop accurate insights/models. It also drives down the productivity of the data team. Data workers spend most of their time trying to piece together existing data to clean it and create better quality. Data scientists report that they spend more than 82% of their work cleaning and preparing data for AI/ML applications.

Organizations need to focus on creating good data pipelines, reliable data flow infrastructure to build good data leading to better accuracy in deriving insights.