Article

    How to make sure your AI project isn’t one of the 80% that fail?

    Most AI projects don’t deliver. Let’s change that.

    Service

    Data and AI

    Industry

    Data and AI

    There are three non-negotiables when moving towards AI agents and enterprise AI: clear data strategy, effective data governance, and a robust technology platform.

    More than 80% of AI projects fail, marking twice the general failure rate of IT projects. This striking statistic is quoted by many expert organizations, including a global non-profit think tank, The RAND Corporation.  

    The biggest reason AI undertakings fall short is the underlying data quality. Missing or bad data has been studied as the root cause of failure in over 70% of AI projects. This is especially common in the “enterprise AI” projects that aim to use enterprise-wide data efficiently, as discussed in our previous blog post. 

    AI systems do not live up to their promise without high-quality, relevant, and well-governed data. To put it more eloquently, it’s garbage in, garbage out. Making data quality–and quantity–a priority is a pivotal factor in the success of AI projects and on the road towards autonomous AI agents, i.e., AI systems that solve problems and operate independently. Other reasons for AI projects missing their mark are misunderstanding the problem domain, misaligning business goals, and over-promising what AI can achieve. AI is not a silver bullet that can fix bad management or missing objectives and strategy. 

    When tackling these most common causes of failure on your AI journey, three interlinked concepts will come into play: a data and AI strategy, a data governance model guided by that strategy, and a robust data platform. 

    Data governance models help realize the strategy and vision while maintaining data integrity, security, and compliance, and providing a structured approach to managing data assets. On the other hand, building a solid data platform ensures that data is well-organized, accessible, and reliable. These elements together form a strong foundation supporting AI solutions' development and deployment. 

    Purple and pink abstract image of an arrow

    Realizing your data and AI strategy 

    To start with the “Why”, you will first need a clear data management vision and strategy. Top management commitment is mandatory to ensure all stakeholders work towards the same goals. Essential building blocks of an actionable data and AI strategy also include metrics that must be unambiguous, transparent, and effectively measurable. 

    A data governance model is the “How” of realizing the set data vision and following the strategy. A proper data governance model helps avoid silos and communication breakdowns between teams. It enables centralized security and privacy governance and increases productivity by accelerating the data and AI governance processes related to security and privacy gates.  

    A well-defined data governance model defines the ways of working related to data, how information is managed to stay compliant, and standardized terminology so people speak the same language. It describes the standard operating procedures related to data with KPIs and processes that are followed and steered accordingly on all levels of the organization.  

    A vital part of data governance is also the “Who”, i.e., the organization and roles. Efficient organization and team structures, where the needed competence is identified at all levels, help respond to the lightning-quick pace of change. It also enables both in-house and partner people to stay on top of their game.  

    Finally, a data governance model defines the lifecycle processes for data, ensuring that relevant data is stored correctly, and unnecessary data gets scrapped properly. Without such lifecycle governance, your technical platform will include a lot of garbage, making your beautiful data lake look more like a data swamp in no time.

    Building technical groundwork on leading platforms 

    A data platform, a centralized repository for different data stages and products, ensures that information is accessible, usable, safe, and trusted. This reduces the risk of missing data and improves the data quality used in AI models. Data integration from various sources enables a more comprehensive view of the data. It reduces direct dependencies between systems, helping IT lifecycle management and identifying data quality issues early in the project lifecycle. Identifying the best AI use case candidates becomes easier when accurate data is available. Also, AI use-case projects become faster, simpler, smaller, and cheaper as they can rely on a solid foundation instead of developing ad-hoc data pipelines and point-to-point integrations. 

    AI poses new requirements for data platforms and solutions. The computation power needed to run Generative AI models can increase by a factor of ten or even a hundred from what the organization used to need. Fully on-premises-based data warehouse solutions are rarely equipped to run AI models efficiently, as they are complex to set up and lack the features and ability to scale up and down quickly. Older data warehouses usually concentrate on cold data (as opposed to real-time) visualization for decision support, not optimizing processes and raising automation levels. Therefore, if you do not have a proper public cloud platform (such as Microsoft Azure, Amazon Web Services or Google Cloud Platform) set up for your organization, it is time to take that leap. 

    There are several data platform products on the market, with modern ones usually available as PaaS or SaaS (Platform/Software as a Service). On Nortal customer projects, we have worked with various platforms like Microsoft Fabric, Databricks, and Snowflake. While every platform has its pros and cons, the spotlight is currently on Fabric, Microsoft’s new SaaS data platform.  

    Microsoft Fabric boosts productivity by providing unified user experience and seamless integration with various services, reducing system dependencies and offering a holistic view of data. Developers can leverage familiar tools and technologies, speeding up the development and deployment of analytics solutions. The platform also accommodates modern data management patterns like data mesh and medallion architecture. 

    For data governance, Fabric includes robust features such as a centralized admin portal, data lineage, impact analysis, and advanced information protection and auditing capabilities. Additionally, Fabric integrates deeply with AI services, offering prebuilt AI models and tools. A collaborative environment for building, training, and deploying AI models is provided, making AI accessible to data scientists and business users alike.

    A pink and purple abstract image

    Our impact on leading industrial companies’ AI journeys 

    By building solid foundations for data management, Nortal has empowered our industry-leading clients to achieve their data and AI goals and realize tangible business benefits.  

    Outokumpu, a global leader in stainless steel manufacturing, faced challenges in gathering data from scattered sources and further utilizing it effectively. We have helped Outokumpu to integrate data from various sources, including business applications, production systems, and IoT sensors. This has enabled them to automate previously fully manual processes, such as coil surface inspection quality control. A comprehensive digital platform allows Outokumpu to develop AI models for production optimization and prepare AI use cases related to other core business processes. Please read more about our data platform customer cases in our previous blog post. 

    Another global manufacturing client of Nortal is focused on enhancing its digital capabilities on the shop floor. The company aimed to implement a standardized digital shopfloor platform across numerous sites, leveraging Microsoft Azure and Fabric to provide connectivity, data analytics, and visualization capabilities. The jointly created platform democratizes data access from machine operators to executives, enabling real-time AI-based insights and continuous improvement. 

    Nortal: Your partner in building the foundation for Enterprise AI 

    Are you happy with how your digital transformation and AI journey is going? Or are you in the four out of five organizations whose AI projects turned out to be a disappointment? 

    Applying AI is imperative in today’s business landscape, but applying enterprise AI and moving to autonomous AI agents will be a game-changer. The journey from personal AI assistants to corporate-wide use of autonomous agents is exciting and promising to revolutionize business processes and decision-making. However, there is no way around it: a solid data strategy, governance model, and platform are crucial milestones on the roadmap.  

    Establishing a well-designed data governance model helps maintain data integrity, security, and compliance, providing a structured approach to managing data. With a comprehensive data platform built on solid technology foundations, organizations can ensure that their data is well-organized, accessible, and reliable. When steered by a clear data strategy and vision, these elements together create a strong foundation supporting AI solutions' development and deployment. These steps will enable you to move your AI projects from the 80% failure bucket to enabling business innovation and competitive advantage. 

    Get in touch

    Nortal is a strategic innovation and technology company with an unparalleled track-record of delivering successful transformation projects over 20 years.