The Wild West of Data Governance

Jul 9, 2025

6 min read

Part two of a three-part series

While the quantity and access to data have increased, with Gen-AI playing a significant role, no one is sure of the rules. Can I ask ChatGPT to write a template for my company’s SOPs - probably not? Can I plug my internal data into an LLM? It is the Wild West, and the consequences of misusing data could be make or break for the regulated industries. I spoke to Lynn Epstein, BlueberryLabs COO, and Samuel Fryer, BlueberryLabs CEO, to get a handle on the Wild West and how organizations might navigate it.

So much data, so little room for mistakes

In regulated industries, the wrong data inputs have serious consequences. I ask Sam to explain this in the context of the changing nature of data governance: “Data relevance has naturally merged with the data governance conversation. When you talk about data relevance, you also have to have data governance principles established, meaning you have the right compliance to use the data. Now, knowing how you can store data and where the original article or source is becomes critical. And that’s all before you start using large language models.”

“It’s a new era and inevitably, new questions are being asked. We have top-level law firms that are chomping at the bit to leverage AI tools as part of their processes, but they must first ask themselves how they maintain compliance. This is especially important in a world where a misplaced word can completely alter meaning and have legal ramifications.”

Compliance and consternation

Managing data in the age of AI requires more conversation

As we enter a new era of data explosion, I wanted to understand how companies are grappling to stay ahead and remain compliant. Sam says: “GDPR is a good reference point because I still remember those emotions of being Chief Revenue Officer and marketing sat under my remit. I remember spending days reviewing policies and thinking about them. What did it mean? What was our legitimate interest? Even though you have been operating for a long time and have a very clearly defined target customer, it’s easy to start to question yourself.”

“Then come the AI-specific questions: can we train models with this data? Are we allowed to summarize the data, and if so, what can we summarize? I think that over time we will get the answers.”

“Technology is moving so fast. Things like ChatGPT enable people to run prompts across masses of information. One of the ways to tackle this is to always reference the original source. Allow people to know where it came from. People get into trouble when they start trying to steal data and say, “I created this”, when in fact they took it. It's the same when I was at university, if you're writing a dissertation, and if I didn't reference the original source, I would get marked down. If you did it too many times, you failed the course. The same principles apply to using information generated from AI - show where the original idea or data came from, and you can’t go far wrong.”

Lynn adds: “You go to any person now in knowledge management and they're usually more frightened to do anything different or new because they're afraid of being told later that it wasn’t compliant. To that end, they are putting it back on the technology company to say, "Indemnify me”. “Tell me it's okay that I’m using the output of Gen-AI for X, Y, or Z - tell me that this output or data is compliant.”

“Of course, you can verify the data compliance of data providers, but you have to take ownership of your own data governance.

It’s no good to rely on other corporations to tell you whether or not you’re compliant. Returning to the drawing board and reminding yourself of good data governance principles is a good place to start.”

“For example, having a robust method to ensure data integrity is critical when bringing AI into an organization. Failing to structure and cleanse your data can result in poor business decision-making [1] and in regulated industries, the consequences can be far worse. Going back to Sam’s point about the source, your internal data governance processes should allow you to strip the layers back and be able to pinpoint how the technology has gathered or summarized its output, and have controls and policies in place to reduce the risk of data poisoning, or inappropriate data sets being ingested by large language models.

With US states starting to draft AI-specific legislation, in addition to the EU’s AI Act, the requirement for what Dr. TJ Jiang, co-founder of AvePoint and data governance expert, describes as "model-agnostic solutions to manage compliance to avoid consequences from regulators" [2]. This is no easy task, but working with experts like us is a great way to build on these foundations and get AI ready.”

Do you have the right tools for the job?

With so many companies using the phrase ‘AI’ in their marketing materials, it’s easy to forget the problem(s) you’re solving, as Sam explains: “Everyone's at least played with or implemented some form of AI into their organization. You shouldn't implement for the hype. Today, we're seeing startups come out every single week claiming to solve a problem with AI. In reality, they've used some shortcuts to do some coding to build something that may be useful on the surface, but struggles to scale and meet the requirements of large enterprise organisations”.

“For these large enterprises, there will be security concerns for any AI-related tool - Lynn and I have lived it. When building BlueberryLabs, we focused on our team's deep domain expertise, from 20 years of experience, working with regulated industries, knowledge management, and competitive intelligence. We have the luxury of looking at all the noise that's in the market, the tools, the implementation of AI, and help people drive the best value inside of an organization and leverage it to their advantage.”

“I can guarantee you almost every company will be leveraging AI, and those who do not will cease to exist. The real question that businesses should now be asking themselves is not simply should they be implementing AI. They must ask what drives the biggest value inside my organization to create efficiency and effectiveness gains, because ultimately, time equals money. It’s a cliche, but the result is more time spent on work that delivers greater value that spent on tasks AI can take care of easily.”

“If I'm a knowledge worker, I can use my expertise to provide greater insight inside my organization rather than having to search for and retrieve data. I can pinpoint specific pieces of content or points in a research paper in seconds while letting AI do some of the other mundane tasks.”

Navigating the hurdles of data governance might mean you have to go back to the start

Going back to the start, where data governance really starts

The pressure of keeping abreast of technological change is as simple as going back to basics and prioritizing the time to audit your organizational data, as Sam explains:

“The core of AI is driven by good-quality data. You need to go across an organization’s core functions to understand how they use data in their day-to-day. To create a central AI-driven knowledge management system, you must understand the data, and consolidate it to leverage it to its full potential.”

“Starting with the data layer will also help you identify the broad range of data sets contained within your organization, and build data quality principles that help improve output but also inform your data governance policies and procedures.

If you don’t understand where your organization’s data comes from and how it was collected, you are crossing your fingers and hoping for the best - that’s not an acceptable approach in today’s business environment.”

A key takeaway from this discussion is that business leaders need to take proactive steps to focus on data ownership and transparency within their organizations - that’s where a key competitive advantage is derived from. If you aren’t on top of where your data comes from, how you got hold of it, and how users inside your organisation consume it, it’s time to head back to the drawing board (your data layer) before you start bringing in AI tools for things like research. Stay tuned for part 3, where we laser in on the importance of data accuracy and how AI can play its part.