Digital transformation is not about IT, but about the data management behind it
MT/Sprout wrote this week about a pattern we at CivAI encounter continually. Organisations want to implement AI, but discover that the foundations for doing so are not yet in place. Data is scattered across systems, definitions vary, and no one can say exactly who owns which data. According to the research Ewald Smits refers to, 63 percent of respondents rate their own data quality as insufficient or moderate, and 73 percent are not yet ready to truly work in a data-driven way.
The challenge is organisational, not technical
The article flips an important switch. Digital transformation is often presented as an IT project, but the most painful bottlenecks are organisational. Who decides if several departments use the same data differently? Who owns a customer file that runs through three systems? As long as these questions remain unanswered, a new tool will be of little help. The CIO is therefore shifting from technical administrator to someone who must guide change throughout the entire organisation.
Why this affects AI even more
With a standard IT system, you sometimes only notice sloppy data at a later stage. With AI, you notice it immediately. A model that relies on conflicting definitions or unclear ownership produces unreliable outcomes, leading to wrong decisions or extra manual work that negates the entire benefit of AI. We also see this in practice with clients of EduGPT, GovGPT, OrgGPT, and CareGPT. The question of whether the AI can do this for us can usually only be answered after clarifying where the data is located, who manages it, and why the same fields are called something different in two departments.
Why we are not alone in this
This is precisely the domain where we consciously choose to collaborate rather than build everything ourselves. CiviQs, founded by Daniel Verloop, mainly helps public organisations with AI governance and compliance around the AI regulation, and has also written about data quality as the Achilles’ heel of AI projects in municipalities. HowSmart.ai helps organisations make their existing knowledge and content more usable with AI. Both partners address a different aspect of the same problem described by MT/Sprout. First get the data management and ownership in order, then add the AI layer on top.
Working together
Do you recognise the pattern from this article in your own organisation? Don’t start with the AI, but with the question of who owns what and whether definitions match. Are you getting stuck with this, or would you like to discuss how the interplay between data management and AI might look? Get in touch via info@civai.eu. We are happy to think along with you, and will involve CiviQs or HowSmart.ai where necessary.
Want to read more? Check out the article on MT/Sprout. You can read more about data quality in AI projects at CiviQs. More about knowledge and content with AI can be found at HowSmart.ai.
Image for illustration, created with AI