Introducing The Capability Question: What AI Enablement Actually Does to People
The adoption metrics look fine. That is not the same as knowing whether your people are getting better.
Organizations are spending significantly on AI enablement and measuring almost none of what matters.
That is not an indictment of the tools. Most of the tools are genuinely useful. It is an indictment of the evaluation infrastructure, which in most organizations consists of adoption dashboards, utilization reports, and the occasional satisfaction survey, none of which were designed to answer the question that actually matters: are the people using these tools more capable than they were before?
Here is what the gap looks like in practice. A company deploys an AI assistant for its workforce. Adoption climbs. Usage metrics look strong. Someone presents a slide showing time saved per week. Leadership is pleased. Eighteen months later the same employees are faster at producing outputs and no better at evaluating them. Nobody measured that. Nobody designed a study to catch it. The capability question was never formally asked.
This is not an edge case. It is the default state of AI enablement measurement right now.
This series is specifically about AI tools deployed to employees and advisors, the enablement context, rather than consumer AI products, which was the territory of the first series on Ground Truth. When a consumer product fails to develop genuine capability in its users, individuals are affected. When an enterprise enablement program fails to develop genuine capability in a workforce, organizations make systematically worse decisions at scale and rarely trace the problem back to its source.
The first two series on Ground Truth diagnosed failure modes in AI product practice and the measurement traps that undermine strategy in research and insights functions. The through line across both was consistent: organizations optimize for the appearance of rigor rather than the quality of the decisions their systems are supposed to support. Series three goes somewhere specific with that diagnosis.
AI enablement is the context where this failure is most consequential and least examined. When you deploy AI tools to your workforce, you are making a bet that those tools will make your people better at their jobs, not just faster, not just more confident, but genuinely more capable of doing the work that matters. Most organizations have no infrastructure for knowing whether that bet is paying off. They have engagement data. They have throughput data. They have anecdotes from the rollout team.
What they do not have is a capability measurement practice.
That is what this series is about. Not whether AI enablement is worth doing. It is. The question is how you build the evaluation infrastructure to know whether it is working in the ways that actually matter, and what you do when the answer is more complicated than the adoption dashboard suggests.
There is also a specific wrinkle worth naming early. AI tools reliably increase confidence. That is one of their most consistent effects and one of their most underexamined risks. Confidence and capability are not the same thing. An employee who feels more certain because an AI system gave them a fluent, authoritative-sounding answer is not necessarily an employee who made a better decision. Designing measurement systems that can tell the difference is harder than it sounds and more important than most organizations currently believe.
The Series: Four Parts
Part 1: What Does Capable Even Mean?
Before you can measure capability you have to define it, and most organizations have not done that work. This part builds a working definition of capability that is specific enough to be measurable and honest about what current AI tools can and cannot actually develop in the people who use them.
Part 2: The Confidence Problem
AI tools reliably produce confident outputs. Employees who interact with confident outputs tend to feel more confident themselves. Neither of those things is the same as being more capable. This part examines the confidence-capability gap, why it is structurally baked into how most AI tools are built, and what it means for organizations relying on RAG-based systems to enable their workforce.
Part 3: What a Rigorous Study Actually Looks Like
If you wanted to know whether your AI enablement program was actually developing capability in your workforce, what would you measure, how would you design the study, and what would you do with what you found? This part builds that framework and connects to a working prototype for practitioners who want to apply it.
Part 4: Toward a Capability Measurement Practice
Evaluation should not be a one-time study run at rollout. It should be infrastructure. This part makes the case for building ongoing capability measurement into AI enablement programs and offers a practical starting point for research, people analytics, and HR leaders trying to get there.
This series is for people analytics leaders, experience researchers, HR strategists, and anyone building or evaluating AI enablement programs who has started to suspect that the metrics they have are not telling them what they need to know.
It is also for anyone who has sat in a meeting where the adoption numbers looked great and something still felt off.
The capability question is worth asking. Most organizations just have not built the infrastructure to answer it yet.
Start with Part 1: What Does Capable Even Mean?
Part 2: The Confidence Problem
Part 3: What a Rigorous Study Actually Looks Like
Part 4: Toward a Capability Measurement Practice
Previously on Ground Truth: The Measurement Traps That Break Strategy — Series 2 Intro
Ground Truth is written by Shannon Coleman, PhD. Product strategy leader and researcher with a background in psychology and quantitative methods. Her work spans govtech, healthtech, fintech, and consumer product contexts, with particular depth in enterprise AI platforms. Portfolio: shannonlcoleman.com