ADKAR’s Ability phase is about having the skills to implement the change. Knowing isn’t enough; people need to be able to do the work in the new way.

For agent-augmented teams, ability looks different than it did five years ago. The premium skills aren’t typing speed or syntax recall. They’re judgment, prompt design, evaluation, and integration. Here’s what that means in practice.

For Individual Contributors (Engineers, Designers, Analysts)

Judgment under uncertainty.
Agents produce output that’s usually good and sometimes wrong. The ability to quickly assess quality—“this looks right,” “this is off,” “this is dangerous”—without exhaustive verification is critical. That’s pattern recognition, domain knowledge, and calibrated skepticism. It develops with practice and feedback.

Prompt design and iteration.
Ability here means composing clear, constrained prompts; iterating when output misses; and knowing when to add examples, context, or structure. It’s a craft, not a recipe. The best practitioners develop a feel for what the model needs.

Testing and validation.
When the agent generates code, specs, or content, someone has to validate it. Ability means designing tests, running spot checks, and knowing what “good enough” looks like for the task at hand. Automated tests help, but human judgment about what to test and when to ship remains central.

Integration and glue work.
Agents produce artifacts; they don’t (yet) integrate them into the full system. Ability means taking agent output and making it work: wiring it up, handling edge cases, fitting it into the architecture. The “last mile” is still human work.

For Product Managers

Judgment on scope and priority.
When agents can generate more, the bottleneck shifts to deciding what to generate. Ability means evaluating ideas quickly, killing bad ones early, and focusing agent effort on what matters. That’s product judgment—sharpened, not replaced.

Prompt design for specs and stories.
PMs who can direct agents to produce useful specs, user stories, and acceptance criteria get leverage. Ability here overlaps with engineering prompt design but applied to product artifacts.

Evaluation of agent-generated product content.
Agents can draft user research summaries, competitive analyses, and feature descriptions. Ability means knowing when the output is usable vs. generic vs. wrong. Same validation muscle as engineering, different domain.

Orchestration across the pipeline.
PMs who understand the agentic workflow—who does what, when to hand off, when to intervene—can design better processes. Ability is system thinking: seeing the full flow and optimizing for coherence.

Developing Ability

Ability isn’t conferred by training alone. It’s built through practice: use the agent, get feedback, refine. The role of the org is to create low-stakes opportunities to practice—pilots, sandboxes, pair sessions—and to recognize and reward ability when it emerges. The people who get good at this become the new craft experts. Invest in them.