Feeding Accurate Data Alone Doesn’t Guarantee Accurate Outcome — Stibo Systems Chief Product Officer

(US & Canada) Neda Nia, Chief Product Officer at Stibo Systems, speaks with Mark Johnson, Chief Growth Officer at CoStrategix, in a video interview about the aspects of AI governance and its massive scope, legislation around AI, how Stibo Systems helps enterprises tackle AI governance, the three customer categories, and how Stibo Systems is dedicated to aiding them based on their needs.

Stibo Systems is a privately held subsidiary of the Stibo A/S group, headquartered in Aarhus, Denmark. Stibo Systems’ enterprise master data management solutions bring disconnected data sources together.

Despite the rapid evolution in the AI space, it is a must for organizations to prioritize requirements concerning risks and liabilities, says Nia. She adds that CEOs, Chief Security Officers, and CDOs must continue emphasizing risk mitigation in any decisions taken, especially around adoptions, given the associated unknowns.

Further, Nia states that AI governance comes into play in the concept of risk and liabilities, which include ethical, transparent, and accountable development and use of AI systems. AI governance also plays a part in legislation, wherein companies need to stay compliant with the brewing potential regulations, she says.

Besides, AI governance also involves revamping people and processes through training and education on ethical practices, says Nia. Apart from that, organizations must be aware of the happenings in the organizational data landscape, assess privacy measures, and consider sustainability considerations.

According to Nia, the computation needed for a certain algorithm is now way higher than before, which impacts sustainability goals. Considering Stibo Systems’ expertise in data, tackling AI governance from the data perspective has been helpful, she adds.

Moving forward, Nia zones in on the aspect of legislation around AI and shares the example of one of the newly formed U.S. legislations called the Algorithm Accountability Act. She continues that its enforcement would mandate companies to conduct impact analyses or assessments of decisions made automatically by AI.

This act could evaluate potential impacts on aspects of accuracy, fairness, bias, discrimination, privacy, and security, says Nia. She maintains that given the massive scope of the act, a single tool cannot do everything, but a platform like Stibo Systems will provide customers with secure control over enterprise data.

Delving further, Nia shares that customer enterprises can assign experts and authorize them to take actions, address what makes them non-compliant from a data perspective, and then pick a component.

For instance, she says, if expert users pick accuracy as a component that they want to manage on Stibo’s platform, they will have to define what accurate data is. After giving the data definition to the tool, experts need to put governance around model training so that the machine learning models and algorithms only have access to high-quality, accurate data.

Even after the model creates an output, experts can add further governance to compare whether or not the result is based on the accurate data that was fed. Further, she asserts having a human in the loop to fine-tune the model or findings, which is referred to as supervised learning.

Thereafter, Nia believes that while more automation may take place, feeding accurate data alone does not guarantee an accurate outcome. Putting in governance is critical, and Stibo Systems’ tool helps with that.

Furthermore, Nia reiterates the three customer profiles of makers, shapers, and takers and states that Stibo Systems focuses on meeting them where they are. Elaborating, she says, the makers pioneer in building their own LLMs, while shapers prefer to take the available tool and add their own data and algorithm to that.

The third category that Stibo Systems works with is the takers who interact with topics that are tried and tested in the market. She opines that it may be due to the nature of the business, being in a highly regulated sector, or the organizational culture.

Nia states that each of the customer profiles has specific business requirements, and they need to deal with potential risks in a way that makes sense for their organization. Stibo Systems is driven to meet customers where they are and where their data is to be able to help organizations achieve business goals.

Concluding, Nia acknowledges the complexity of the topic and the need for organizations to take an iterative approach to solidify the AI strategy. She affirms that Stibo Systems will aid customer enterprises with high-quality data and by governing the process along the way.

CDO Magazine appreciates Neda Nia for sharing her insights with our global community.