Artificial intelligence (AI) continues to drive discussions across the enterprise, from the board to the workforce. C-suites are expected to drive discussions and strategies on how to best tap the technology before the competition does. Gartner warns, however, that AI poses unique challenges for governance that must be directly addressed to mitigate risk and take advantage of opportunities.
Heather Gentile, executive director of product and management at IBM, discusses how organisations develop AI governance and the benefits and challenges they experience around it.
Importance of AI governance
Gentile says that AI governance involves directing, monitoring, and managing the AI activities of an organisation. Moreover, she says an informed AI governance strategy includes ethical considerations like explainability, fairness, robustness, transparency, and privacy to guide responsible AI adoption.
According to the IBM executive, building an AI governance strategy can help organisations adopt AI at scale while proactively mitigating risks.
“AI has a profound opportunity, but it poses a global challenge. Analysts project AI to unlock an astounding 16 trillion in value, helping to address some of our most pressing challenges in healthcare, manufacturing food, production, climate change, and several others. However, it can only meet its potential if people trust the technology as a force for good,” Gentile says.
As innovations with AI increase, governing the technology has been a recent focus for organisations to support technological initiatives with a strong AI governance program.
“What we see changing globally is a shift for governance to be more strategic across the enterprise with the power and innovation opportunity of generative AI along with the costs associated with supporting large and powerful models,” she posits.
Pursuing an enterprise-grade AI governance program
Gentile says it is important to involve stakeholders from the organisation with diverse perspectives to establish an enterprise-grade AI governance program, including the CEO for organisational accountability, CIO for perspective on information technology, and the CISO for information security. Gentile adds that there is also a growing involvement with the chief financial officer to assess profitability, the marketing department brand risks, and the human resources department for employee accountability. Aside from that, the chief risk officer can provide perspective on the risk to the enterprise and compliance, and the chief privacy officer and the chief data officer for privacy and trust in data operations.
“It is these perspectives of the individuals that create an informed view of the consideration that needs to support an effective AI governance program for the organisation to adopt AI responsibly. The program must consider how to support an adoption that drives innovation and business opportunity. It also needs to mitigate risk and align to ethical standards,” she says.
Gentile reminds an AI governance framework must consider the core values and priorities of the business.
“I think one of the greatest common mistakes organisations may have in trying to roll out their AI governance program is manual processes that can be inefficient in subject to human error. The best practice is for responsible AI adoption to leverage technology, wherever possible, to standardise the process and push out best practices across the enterprise,” she says.
Gentile says that AI governance needs to become part of the culture of an organisation through investing in employee training and updating policies and procedures to understand the risks involved.
She also underscored the importance of fixing data before deploying an AI governance program, or it can run the risk of becoming ineffective.
“The models are only as good as the data they are trained on. You cannot successfully adopt AI or be confident in the results the model is returning without good data governance and confidence in data,” shares Gentile.
At IBM, we have a saying that there is no AI without IA (information architecture)
Heather Gentile
AI governance program IBM watsonx
To enhance compliance for businesses to maximise the full potential of AI, IBM built watsonx, an end-to-end automated governance solution which provides an integrated platform supporting three pieces of AI adoption: Watsonx.ai, watsonx.data, and watsonx.governance.
Watsonx.ai workbench can facilitate the building, deploying, testing, training, and tuning of AI models. The 2nd piece watsonx.data, is a data lake house that provides secure access to data across any hybrid cloud. Finally, watsonx.governance wraps around the platform, helping businesses manage their AI responsibly across the enterprise at scale, manage risks, and ensure compliance across the lifecycle.
“I think our solution is unique because not only do we provide governance as part of this integrated watsonx platform. It is an open toolkit that supports AI anywhere. It is capable of covering both predictive ML and generative AI models, including 3rd party and open-source models on any AI application,” she explains.
Interested organisations can visit IBM’s website for free toolkits, information about the technology, the ability to sign up for a demonstration and create a trial account, and experiment with watsonx.
“We also collaborate with businesses across different industries on different proof of concept programs. Those are free engagements where we will talk with an organisation about where they are in their AI governance maturity and what they are looking to do to get to the next step. We will come up with some objectives in use cases and collaborate," she adds.
Advice for C-suite
“It is a very exciting time for AI overall,” Gentile says.
As the C-suite establishes accountability, Gentile reminds the importance of being educated on the latest industry trends, opportunities, and risks. She also reiterates how critical ethics is in setting the strategy and that organisations must ensure the AI governance program reflects it. She advises C-suites to think about how they are establishing explainability and transparency and their controls around privacy.
“These are all important things that are not only going to protect your organisation but also help you to meet the regulatory requirements,” Gentile says.
Click on the PodChat player to hear in detail Gentile’s perspective on pursuing an AI governance strategy for organisations:
1. What is AI governance?
2. What has given rise to AI governance?
3. Why has AI governance become important in 2024?
4. What is involved/included in AI governance?
a. What are the critical components of enterprise-grade AI governance?
5. Are all AI governance approaches equal? (i.e., can enterprises use the same template to build their AI governance?
6. Who should be involved in drafting/managing an organisation's AI governance strategy?
7. Can you highlight some common mistakes organisations might make in the pursuit of an AI governance strategy?
8. Can you cite notable learnings for C-suite leaders and the Board in the pursuit of an AI governance strategy?
