Generative AI tools can draft reports, analyze complex datasets, and compose polished client emails in mere seconds.
But sheer speed does not equal true productivity if the output lacks strategic nuance, audience awareness, or sound professional judgment.
In their Harvard Business Review article, “Help Employees Get Better-Not Just Faster-with AI,” strategy consultants David S. Duncan and Tyler Anderson explain that while AI possesses vast global knowledge, it has zero context regarding your organization’s internal politics, specific client histories, or non-public market realities.
True productivity in the AI era comes from mastering a new hybrid skill: making your tacit human judgment explicit enough to evaluate, guide, and refine artificial intelligence.
To ensure AI enhances the quality and accuracy of your work rather than merely accelerating generic first drafts, Duncan and Anderson recommend a systematic four-step workflow:
- Establish an Initial Point of View: Before opening any AI tool, scope your task and form a preliminary hypothesis. Define what specific question you are answering, who the audience is, and what a strong output looks like. Having an initial baseline gives you an objective foundation to evaluate AI outputs. Without your own baseline, you have no basis for critiquing the machine’s work.
- Collaborate Across Multiple Modes: Most professionals only interact with AI in a single mode: generation. To unlock higher quality, push the tool across four additional modes: ask it to critique its own weak assumptions, compare trade-offs between different options, simulate how key stakeholders might react, and challenge the reliability of its data sources.
- Analyze the Differences: Compare your initial baseline hypothesis against the AI’s output. Carefully categorize the differences into what AI added that you missed, what it got blatantly wrong, and most importantly, what looks right but isn’t. Plausible-sounding errors grounded in generic industry assumptions represent the costliest AI traps, and catching them requires genuine domain expertise.
- Deliver the Output with a “Reasoning Trail”: When submitting deliverables to team leads or clients, include a brief explanation detailing what AI initially generated, what you changed, and why. Documenting where AI succeeded and where it struggled makes your judgment visible, making your reasoning coachable and sharpening your expertise over time.
In knowledge work, AI models can easily generate 90% of a draft, but the entire value of the final product lives in the remaining 10%—the human editing, contextual correction, and strategic alignment applied on top. By transforming from a passive AI user into an active editor and calibrator, you ensure that technology makes you genuinely better at your craft, rather than just faster.
*Ideas for this blog taken from: Duncan, D. and Anderson, T. “Help Employees Get Better-Not Just Faster-with AI,” Harvard Business Review online, June 15, 2026.
