AI tools and governance
One participant in the roundtable expressed that CDOs have found themselves in a position with more leverage in the boardroom which means they not only have to educate business leaders on AI, but also evangelise a data narrative in a way that impacts business areas that previously did not engage with data. One of the most important parts of this approach is to ensure the seriousness of data governance is understood and a core section of data literacy upskilling.
Another participant from an insurer highlighted their experience with shadow IT and the impacts it had had on governance and compliance in a highly regulated industry. In prior years, the conversations from data leaders have been about getting non-data professionals to appreciate the issues that surround shadow IT and silos, whereas now the conversation is shifting into ensuring ethics, governance, and compliance are consistently monitored.
However, governance is not always a net benefit to a business and can be a bottleneck. As one roundtable participant explained, their business had its innovation hindered by governance three years into a five-year plan. The data team was forced to backtrack some of its work in order to review any new data product which drastically impacted innovation and any momentum built up for other data initiatives.
The representative explained how their team had to go back and forth between their team, the board, and another part of the company. The board believed they were developing similar enough data products that meant each item would not need to be individually approved, but this was not the case. Now, with new items having to be reviewed, the review board has been regularly blocking developments due to governance issues.
Governance acceptance and championing
Without buy in from the top tier of a business, governance will always be an issue. As people are given ownership of different data products, their enthusiasm and appreciation for why governance is required is essential to maintaining quality, compliance, and accessibility – as soon as governance gets undermined, a snowball effect of issues can take place. This was highlighted by one participant who described their difficulties surrounding ownership for a data mesh set up that is being implemented within the business.
That being said, there is not a one-size-fits-all solution to governance. One DataIQ member discussed how their team has gone through a journey with data product governance, where they are trying to focus governance on adding value rather than a singular approach. Having different levels of governance based on the style of data product has helped ensure whatever needs to be highly governed is indeed governed, without stifling innovation. This does require more time and consideration when compared to a one-size-fits-all solution, but the benefits are unsurpassed and provides huge levels of stability moving forward.
To ensure governance is accepted and appropriately championed, governance must be given a platform at the top end discussions. It was noted by numerous roundtable participants that data governance is often separated from the business – described as being “black boxed” – which prevented it from easily being integrated into day-to-day activities.
Additionally, this makes developing workflows with governance principles built in has been difficult to perfect. One data leader explained that they tackled this within their team by actively bringing governance into the conversation. For example, when having conversations with decision makers about new data products, one team physically brought the governance officer to all conversations to make governance a part of the conversation every time by default.
GenAI’s impact on governance frameworks
With the rapid growth of generative AI (genAI), there have predictably – and correctly – been amends to existing governance frameworks as more people pay attention to data. Businesses that are embracing genAI technologies need to ensure existing governance practices are visible and understood by non-data departments as they will be the ones benefiting from the genAI tools. One roundtable participant explained how their team ensures any AI workflows go through the CIO for approval to maintain governance and increase transparency.
A number of teams mentioned they were caught in a struggle to separate out the differences required between established AI data governance and the new genAI governance. Once way around this hurdle raised by one member was to create a parallel governance framework that is specific to genAI and this included separate governance stories and review boards. Of course, this is more resource intensive, but the level of transparency and a clear distinction between the types of AI can be incredibly beneficial to non-data teams.
One area of concern was that as there is an increase in vendor tools to introduce AI into their products there is a higher chance of shadow AI technologies being introduced. One participant mentioned that they have stopped signing longer contracts to be able to negotiate this particular aspect and hold their vendors to a higher standard of transparency regarding AI workflows. It was noted that numerous members had experienced issues with vendors being unable to answer what happens to the data within these new tools and that terms are not being updated to reflect the new features.
There was also a discussion around the balancing act between gaining efficiencies across organisations with new genAI tools and keeping risks low – for example, using genAI tools for copy. It was unanimously agreed that whatever the end use of the genAI tools, the levelling of automation to risk was of pivotal importance. Solutions include maintaining a human in the process and reviewing everything created before it reaches the general population.
Ultimately, the fact is that genAI and AI tools have become a daily part of data as a business, and this is showing no sign of slowing down. It is pivotal that data leaders and their teams update their governance frameworks to ensure continued success, compliance, and improved transparency across an organisation.
Get in touch with the dbt Labs team to discuss your governance needs.
To get involved with upcoming exclusive DataIQ roundtable discussions, click here.
