Data Booster
4 min read
October 5, 2022
Ask five organizations what they mean by data literacy and you will get five answers, and at least two of them are describing data fluency. The difference is not academic once you start writing a program: it decides who you train, to what level, and how you will know whether it worked.
Borrow the language analogy. Literacy is the ability to read and write. Fluency is the ability to do the same easily, accurately and at speed. Applied to data: data literacy is the ability to read, write, communicate and reason with data. Data fluency is the ability to do all of that without friction, in the flow of daily work. Fluency is not a different skill set — it is the same skill set at a higher level.
That matters, because most organizations do not need everyone to be fluent. They need a large group to be literate, a smaller group to be fluent, and a handful of specialists to go deeper than either.
Then GenAI arrived and added a third term. AI literacy is not data literacy with a new label. It overlaps, but it adds skills the older definitions never had to cover: knowing when a model is likely to be wrong, judging output you cannot trace, understanding what happens to the data you paste into a prompt, and recognizing where a tool should not be used at all.
Under the EU AI Act, that last group of skills is no longer just good practice. Organizations deploying AI are expected to ensure their people have a sufficient level of AI literacy — which turns a training ambition into something you have to be able to demonstrate.
The practical consequence: data literacy and AI literacy are not sequential. You do not finish one and start the other. Someone who cannot read a chart will not spot a plausible-looking wrong answer from an AI assistant either, which is why the two belong in one program rather than two competing ones.
Start from the capabilities the organization needs, then work back to the skills each group actually uses. Different roles need genuinely different tracks — a track for someone who reads dashboards looks nothing like a track for someone who builds them.
We work with four broad groups: users who work with dashboards, reports and assistants; builders who know a business domain deeply and use data to act on it; specialists who build and maintain the data and the models; and leaders who decide where AI creates value and how risk is managed.
Within each group, both hard and soft skills matter. Writing data leans on hard skills in tools like SQL and Tableau. Communicating with data is mostly soft skills: asking the right question, and turning an answer into something a colleague can act on. Both are teachable, and neither is optional.
Skill levels are more useful than labels. Reading a bar chart correctly is a basic skill. Drawing a defensible causal conclusion from an experiment is an expert one. Naming the level per role, and measuring against it with an assessment before and after, tells you far more than deciding whether to call the program "literacy" or "fluency".
One caveat we run into often: if everyone scores high on your baseline assessment before the program even starts, the assessment is too easy — it is not that the organization is already skilled. A baseline with no headroom cannot prove anything.
And whatever you call it, it does not finish. Tools change, models change, and the level that counted as fluent two years ago is the floor today. The useful question is not which term to use, but whether you are still raising the bar.
Member of
Nominated
