Building a Strong Ethical AI Approach
5 minute read
1g
From Risk to Resilience:
Building a Strong Ethical AI Approach - an overview for Leaders
Generative AI offers incredible potential, but its power is matched by significant risk. An AI tool that generates biased hiring recommendations, leaks sensitive data, or infringes on copyrights can transform from an asset into a major liability overnight. That’s why a “Strong Ethical Approach to AI” (one of the 8 core capabilities of an AI savvy organization) is a strategic imperative for building trust, managing risk, and securing long-term business resilience.
But how does an organization move from having no formal ethical focus around AI to embedding responsible AI into its very culture? This journey unfolds across four distinct maturity levels. Here’s a detailed, step-by-step overview approach to guide your organization’s progression from Level 0 to a GenAI-Savvy Level 3.
Level 0: No Focus – The Unaware Stage
At this stage, ethical considerations for AI are absent. AI tools may be used in an ad hoc, fragmented way with no formal guidelines, oversight, or awareness of the risks involved, such as algorithmic bias, data privacy breaches, environmental impact or IP infringement. Decisions are driven by technical feasibility or financial considerations, without regard for potential ethical consequences or long-term business impact.
How to Move to Level 1:
Your primary goal is toraise awareness, establish a foundational understanding of why ethical AI matters and begin creating momentum. This begins with leadership engagement and initial, focused experiments.
Start Your Own Learning Journey: As a leader, you cannot champion what you don’t understand. Begin by building your own GenAI-savviness. Develop your own AI-Competence plan based on the CLEAR model. Use personal TERA cycles (Trial, Explore, Reflect, Apply) to build your AI-realted leadership abilities and particularly to explore the ethical dilemmas AI presents. Experiment with GenAI tools to understand their limitations, like hallucinations or bias, firsthand. This direct exposure will give you the credibility to lead the conversation.
Launch an "Ethical AI Discovery" Initiative: Frame an initial, low-risk TERA cycle focused on uncovering ethical risks within your own context. Use the LENS model, specifically focusing on Stewardship and Leadership & Strategy.
Example TERA Cycle:
Trial: Run an "Ethical Decision Challenge." Present teams with realistic scenarios, such as an AI system showing bias in performance reviews or a marketing tool generating infringing content. Ask them to discuss the ethical problems and potential solutions.
Explore: Facilitate sessions to collect the insights. What risks did teams identify? What ethical principles did they feel were most important? Where were the biggest knowledge gaps?
Reflect: Synthesize these findings to create a compelling case for why a formal ethical approach is necessary. What are the biggest immediate threats to your organization?
Apply: Use these insights to secure leadership buy-in for creating the first draft of an AI ethics policy. Perhaps turn the TERA cycle into a training modules which can be repeated. Identify supporters and evangelists to spread the message.
Create engagement by repeating TERA cycles, showing visible ethical behavior and supporting followers and evangelists in running their cycles to raise awareness.
Level 1: Initial Exploration – The Awakening Stage
At Level 1, the organization recognizes the importance of ethical AI. Discussions begin, often in response to an early issue or growing public concern. However, efforts are reactive and inconsistent. There may be informal guidelines, but there are no formal governance structures, and mechanisms for transparency or human oversight are largely absent.
How to Move to Level 2:
The goal here is to formalize your approach by establishing clear structures, principles, and initial control mechanisms.
Form a Guiding Coalition: Establish a cross-functional AI Council or Ethics Board. This group should include representatives from legal, IT, HR, business units, and compliance. Their initial mandate is to draft the organization's first formal "Ethical AI Charter," defining core principles like fairness, transparency, accountability, and data privacy.
Run TERA Cycles to Test Governance Mechanisms: Use iterative experiments to build and test your governance. here is an example:
Trial: Select a single, high-impact but low-risk AI use case. Mandate a Human-in-the-Loop (HITL) protocol for this process, requiring human review of all critical AI outputs. Document the entire process, including the criteria for human intervention.
Explore: After a set period, gather data. How often did the human reviewer override the AI? What types of errors were caught? What was the impact on efficiency and trust?
Reflect: Discuss what the trial revealed about the necessity of human oversight. What worked well in the HITL process, and what was cumbersome? How can the process be refined?
Apply: Formalize the HITL protocol as an prerequisite for all similar high-stakes AI applications and share the success story to build support for broader implementation.
Launch Foundational AI Ethics Training: Move beyond informal awareness sessions. Develop and roll out a training module for all employees that covers the organization’s new ethical principles, common risks like bias and hallucinations, and their responsibility in using AI ethically. Here as well, you can use TERA cycles to systematically learn what training works best.
Level 2: Emerging Capability – The Structuring Stage
At this level, a formal ethical governance framework is in place and gaining traction. Clear guidelines exist, roles for AI ethics stewardship are being assigned, and efforts are underway to build transparency into systems. The organization is actively monitoring AI regulations. However, enforcement and integration across all business units may still be inconsistent.
How to Move to Level 3:
The focus now is on embedding and operationalizing your ethical framework, making it a proactive, integral part of the AI lifecycle and the organizational culture.
Integrate Ethics into the AI Lifecycle: Shift from reviewing ethics after the fact to building it in from the start. Ensure that new AI initiatives complete "Ethical Risk Assessments" before they begin . This should include checks for potential bias in training data, transparency requirements, and data privacy compliance. USe TERA cycles to ensure that such processes are meaningfully integrated into operational workflows.
Run Advanced TERA Cycles on ER-AI such as sustainability and resources usage, IP infingement or bias. Here is an example for Explainability (XAI):
Trial: For a key AI system, pilot the use of eXplainable AI (XAI) tools that provide understandable reasoning behind the AI’s decisions. For example, an AI hiring tool should be able to explain why it ranked a particular candidate highly.
Explore: Gather feedback from the teams using these tools. Did the explanations build trust? Did they help identify hidden biases?
Reflect: Assess the value of XAI in fostering accountability and critical thinking. What level of explainability is required for different types of AI applications?
Apply: Based on the findings, establish a policy that requires a specific level of explainability for all high-risk AI systems.
Proactively Monitor and Audit for Bias: Don’t wait for problems to arise. build continuous monitoring systems into processes to audit AI outputs for fairness and discriminatory patterns. Publicly report on your commitment to ethical AI, sharing both progress and challenges to build external trust.
Level 3: GenAI-Savvy – The Embedded Stage
At Level 3, a strong ethical approach is no longer just a policy—it's a core part of the organizational culture. Ethical governance is fully integrated into strategic planning and daily operations. Mechanisms for fairness, transparency, and human oversight are proactively designed into all AI systems. The organization actively anticipates regulatory changes and is recognized as a leader in responsible AI. Most importantly, employees at all levels feel empowered and responsible for upholding ethical standards in their daily work with AI.
At this stage, the journey doesn't end. A GenAI-Savvy organization understands that ethical stewardship is a continuous commitment. It uses the TERA-LENS model not just for transformation, but for ongoing adaptation, constantly refining its practices as technology, regulations, and societal expectations evolve.