By Michael Moulsdale
Published on May 28, 2024
AI is already here, and new applications and advancements show up daily. The technology is expected to impact 40% of all jobs globally. Not office jobs or professions in advanced economies; all jobs. Organizations ready to take advantage of AI’s transformational outcomes will considerably outpace those that aren’t.
So, is your organization ready for AI?
Below, we will explore the promise of AI to reinforce the urgency of becoming AI-ready. We’ll then explore the importance of data to AI success, define AI readiness, list the key features of AI readiness, and detail the requirements necessary to create an AI-ready organization.
AI is pervasive. It recognizes your face to unlock your phone, personalizes the advertisements you see, suggests a chat reply to your best friend’s message, corrects your grammar, plots your route to the grocery store, monitors your payment for fraud, and suggests what to watch after dinner.
In the workplace, AI is already used by HR to screen applicants, Sales to project revenues, Customer Service to answer customer questions, Accounts Payable to process invoices, and Finance to automate reporting. While the 40% data point is shocking enough, take particular note of an additional finding from the same International Monetary Fund report: in advanced economies with mature industries, up to 60% of jobs may be impacted by AI.
The ubiquity of AI will help it reach the estimated global annual economic impact of up to $4.4 trillion according to McKinsey, mainly from productivity gains of up to 70%.
But, as much as AI promises to do, it can’t do anything without data. AI, especially machine learning (ML), deep learning, and generative AI, need trusted, high quality data to learn, interpret, understand, make predictions, and act on new inputs or data.
AI is only as good as the data it uses. An AI model that’s fed poor-quality or inaccurate data will generate poor results. As an extreme example, it’s unlikely that you’ll accurately forecast sales for your manufacturing operations if your AI model was trained on sales data from the hospitality industry. Less extreme but more plausible, your forecasts won’t be accurate if the AI models were trained on dirty, incomplete, inaccurate, or conflicting data from your organization.
When you’re making business decisions, accuracy obviously matters. AI accuracy depends on data. Unfortunately, 29% of organizations have issues with data that will minimize the value they gain from AI and 24% lack trust in their data according to IDC’s Future of Enterprise Resiliency and Spending Survey.
While not quite one-third of organizations have data issues, more than two-thirds (69%) struggle to deliver measurable returns on AI investment according to a Gartner survey of chief data and analytics officers. This points to the importance of people, processes, and technology to data culture maturity. It all matters.
Data leaders know that trusted data is a prerequisite for trusted AI. Trusted data is achieved by building a mature data culture enabled by people, processes, data, and technology.
AI readiness is an organization's preparedness to successfully deploy and use AI to drive outcomes. A strategy is required to direct your AI initiatives, but it takes a combination of people, processes, data, and technology to get the most out of your AI investments.
People with the right skills, attitudes, and training are required to recognize the value of AI, put AI to appropriate use, ensure the right data is used, and have access to the information and expertise needed to find, govern, and understand the data.
Process optimization ensures an enduring, secure, compliant, and usable data environment to put data to work in AI, and trace AI outcomes to data origins. It’s also necessary to manage people, technology, and related policies appropriately as data makes its journey from source to results.
Data, as explained, is essential to AI. But data needs people, processes, and technologies to create a secure and well-governed data intelligence environment that protects private and sensitive information, ensures proper usage, and enables easy search and discovery.
Technology is AI, but organizations must also ensure supporting technologies such as data catalogs are in place to make AI possible.
Let’s drill down into the third bullet: data. IT leaders must prioritize valuable data for AI initiatives, which, according to Gartner, meets these 5 key criteria:
Ethically governed. Clear principles offer data users clear guidelines on how to use data and why. Data leaders must work with executive leadership to establish lighthouse principles for AI that align with organizational values, provide clear direction to data users, and are specific and clear.
Secure. Take appropriate action to ensure your data isn’t accessible to unauthorized users or the LLMs of others unless you want to share it.
Free of bias. Diversity of data sources guards against bias in AI models. Avoid feeding models demographic data from narrow sets of people of similar age, races, and backgrounds, for example.
Enriched. A smaller sample of high-quality, well-tagged, curated data will outperform a larger sample of lower-quality data any day. Ensure your data is tagged with rules, feedback, and business guidance to inform wise consumption.
Accurate. If something seems fishy, investigate! For example, the codes cashiers punch may not correlate to an actual item, but be simply easier to press (as Snowflake’s Principal Data Strategist, Jennifer Belissent, discovered when she drilled down into a spike in sausage consumption for a food service management company which was not what it seemed).
Being AI-ready is synonymous with being data-driven. But simply using data doesn’t help an organization benefit from data. That requires a data culture, where data-driven decision-making thrives and the organization becomes collaboratively intelligent because data is ethically governed and universally accessible.
Consider what it takes to build an AI or ML model, large language model (LLM), or vector database used for Retrieval Augmented Generation (RAG). Data must be found, prepared based on quality, category, and application, traced as it’s used, and recorded to ensure downstream governance and compliance.
A mature data culture using a data intelligence platform ensures data is searchable and discoverable based on AI modeling needs. Developers and data scientists can then find information on data quality and reliability so they make more informed decisions on which data to use and the resulting integrity of their AI models. If they have questions, the data intelligence platform adds shared annotations and directs people to subject matter experts and data stewards to ask questions related to AI model development.
The result is a more AI-ready organization that enables data scientists to find the right data to train AI models that deliver the most impactful outcomes.
Once AI models are made available to the organization, worker confidence is reinforced using the same metadata, lineage, governance and process controls, and other capabilities that the data intelligence platform delivers. If AI performance is ever suspect or ready for improvements, the same data lineage and provenance information can be used to update, replace, or find alternatives to the original AI training data.
Since good data is the prerequisite for functional AI, AI readiness relies on a mature data culture to support successful AI initiatives. AI training data must be easily discoverable, accurate, well-understood, properly governed, and trusted. AI models must use appropriate and high-quality data to learn from and produce accurate predictions. And, everyone along the way must understand how to find and use trusted data. All of those requirements are inherent to a mature data culture.
Data culture maturity refers to how organizations use data to achieve goals and make key decisions. Mature data cultures make data easily accessible, understood, trustworthy, and relied upon. This requires infrastructure and a cultural shift so the organization values evidence and reason over opinion, consensus, or rank. It also sets the groundwork for AI success.
Data culture maturity consists of four core pillars:
Data search & discovery enables users to find, understand, and trust the information they need.
Data leadership aligns data initiatives to outcomes and measures the value to your organization.
Data governance defines how data should be gathered and used via a surrounding structure and support.
Data literacy enhances collaboration and understanding to make data-driven decisions.
A data intelligence platform is a critical component of the AI stack to enable data discovery and data governance and facilitate data leadership and data literacy. This technology centralizes metadata — the data that describes data — from disparate sources to create a unified and automated platform for AI readiness with accessibility to trusted data, governance, and processes.
Preparing data and cataloging it for later search, governance, and usage relies on a data intelligence platform. The platform is then used for classifying data, making more informed decisions about data such as when to archive or how to govern usage, evaluating and recording data quality, detailing data provenance and data lineage, identifying training needs, and more.
Good AI requires good data. Finding, trusting, and using good data requires a mature data culture. In other words, AI readiness and data culture maturity are one in the same.
A data culture maturity model provides qualitative evaluation criteria organizations can use for continuous improvement. As technology and techniques advance, especially in relation to data, the levels of data maturity will also advance and change. Even the most data mature organizations can still find areas of improvement.
With the huge promise of AI, data culture maturity is imperative and urgent. Book a demo with Alation today to learn how you can rise to the challenge.