Safety and AI icons on either end of a wooden board that is balancing on a wooden ball

Making AI real: Balancing speed with safety

Artificial Intelligence

The adoption of artificial intelligence (AI) presents both unparalleled opportunities and significant risks. Rehgan Avon, co-founder and CEO of AlignAI, provided a comprehensive overview of the challenges companies face when integrating AI solutions, including costs, security vulnerabilities, ethical considerations, reputational risks and regulatory requirements, in a recent Risk Series event.

Here are some key takeaways from her presentation:

  • Collaboration and data are vital when implementing an AI strategy
    • Get buy-in from all stakeholders (e.g., leadership, HR, legal, cybersecurity, developers, etc.) and establish common ground and language when talking about AI to ensure everyone is on the same page and understands the purpose and goal of the strategy.
    • Involve organization leaders in the creation of your AI strategy so that it aligns with the organization’s strategy.
    • The data used for AI and machine learning (ML) models and tools must be both high quality and high quantity.

“If collaboration is the #1 obstacle to successful AI implementation, data integrity is #2.”  - Rehgan Avon

  • Consider the ramifications of implementing AI
    • Avon said, “Nothing has gone from a ‘delighter’ to a ‘must-have’ faster than AI,” but just because we can use AI, doesn’t mean we should. It is critical to think through the consequences your organization could face by using AI, e.g., workforce impacts and addressing re-skilling needs if roles will be replaced and the cost to recruit and retain employees with knowledge in AI.

  • Understand the risk of AI
    • Enterprise risk management (ERM) and change management should play a significant role in any company’s AI strategy implementation.
    • AI and ML models are not always interpretable and should not be thought of in the same way as software. “Even if we don’t understand why a model did something legally problematic, we are still liable,” cautioned Avon.
    • The book Weapons of Math Destruction by Cathy O’Neil posits that AI is more dangerous than humans due to mass deployment; if a person makes a mistake with or for a team of 10 or 100 people, that’s one thing, but an AI mistake can potentially affect hundreds of thousands of people.

  • AI best practices are a moving target
    • The industry continues to innovate and develop new AI and ML models, use cases and breakthroughs. The speed of these changes means best practices are continuously evolving.
    • There is a continued need for academia and global governments to define and regulate this industry.