How would you quickly initiate an audit of your AI initiatives to understand their impact on talent in your organization?

Difficult to answer. From an IT auditor's perspective, an audit requires clear parameters to be actionable and therefore useful, even if the items being measured are qualitative.
"Impact on talent" is a nebulous concept. If "talent" means people then what specific factors are in question? Is it something as simple as whether you have gained or lost headcount? Or is it more complex like worker satisfaction or morale changes after each specific AI initiative. Looking deeper, is a morale change due to the AI initiative or the fact that bonuses were just distributed? Did the initiative generate enough revenue to be able to hire high level talent or upskill existing people?
If "talent" means skills, then is the question whether AI initiatives cause catastrophic forgetting preventing resulting in people who can't perform their usual functions without an AI crutch?
Defining which "impact" that is being evaluated will help narrow things down. However, "auditing" an initiative is very challenging since an initiative by definition encompasses multiple processes and can extend out as broad as the overall organizational culture.

Capturing Business Outcomes Through AI KPIs
When driving business use cases for AI adoption, it is critical that outcomes are measured through well-defined Key Performance Indicators (KPIs). These KPIs should fall into three broad categories:

1. Direct Business Impact
This includes metrics such as revenue impact or time savings.
For example, one customer implemented AI in production support. The measurable outcome was a reduction in call resolution duration by 8–10 minutes.
This type of KPI is directly observable, monitored, and reportable, making it easier to quantify ROI.

2. Indirect Business Value
Certain use cases deliver value that is less immediate to measure. A good example is GitHub AI, which generates code, test cases, and improves software quality.
The outcomes here are not captured through a single metric, but instead require tracking over time—such as QA defect rates, production issues, levels of technical debt, and deviations from functional requirements.
While the business benefit is clear, KPIs in this category require ongoing data collection across weekly or fortnightly cycles to provide meaningful insight.

3. Adoption & Sentiment Tracking
In addition to performance metrics, CIOs should track AI adoption sentiment across the workforce.
This can be done through qualitative and quantitative employee surveys, repeated over time, to identify trends in user confidence, productivity, and satisfaction.
Combining sentiment with operational KPIs provides a 360° view of AI’s impact on both business outcomes and employee engagement.