When creating AI governance, differentiate based on risk. High-risk actions, like uploading sensitive company data into a public model, require rigid, enforceable "policies." Lower-risk, judgment-based areas, like when to disclose AI use in an email, are better suited for flexible "guidelines" that allow for autonomy.

The biggest hurdle for enterprise AI adoption is uncertainty. A dedicated "lab" environment allows brands to experiment safely with partners like Microsoft. This lets them pressure-test AI applications, fine-tune models on their data, and build confidence before deploying at scale, addressing fears of losing control over data and brand voice.

Many leaders mistakenly halt AI adoption while waiting for perfect data governance. This is a strategic error. Organizations should immediately identify and implement the hundreds of high-value generative AI use cases that require no access to proprietary data, creating immediate wins while larger data initiatives continue.

Organizations fail when they push teams directly into using AI for business outcomes ("architect mode"). Instead, they must first provide dedicated time and resources for unstructured play ("sandbox mode"). This experimentation phase is essential for building the skills and comfort needed to apply AI effectively to strategic goals.

Treating AI risk management as a final step before launch leads to failure and loss of customer trust. Instead, it must be an integrated, continuous process throughout the entire AI development pipeline, from conception to deployment and iteration, to be effective.

To navigate the high stakes of public sector AI, classify initiatives into low, medium, and high risk. Begin with 'low-hanging fruit' like automating internal backend processes that don't directly face the public. This builds momentum and internal trust before tackling high-risk, citizen-facing applications.

To navigate regulatory hurdles and build user trust, Robinhood deliberately sequenced its AI rollout. It started by providing curated, factual information (e.g., 'why did a stock move?') before attempting to offer personalized advice or recommendations, which have a much higher legal and ethical bar.

When developing AI for sensitive industries like government, anticipate that some customers will be skeptical. Design AI features with clear, non-AI alternatives. This allows you to sell to both "AI excited" and "AI skeptical" jurisdictions, ensuring wider market penetration.

When leadership demands ROI proof before an AI pilot has run, create a simple but compelling business case. Benchmark the exact time and money spent on a current workflow, then present a projected model of the savings after integrating specific AI tools. This tangible forecast makes it easier to secure approval.