You're Using AI at the Execution Layer. The Value Is in the Judgment Layer
Most teams use AI for drafts and summaries — the execution layer. Duane Forrester argues the real leverage (and career moat) is judgment, the thing you can't retrieve.
TL;DR
- Duane Forrester (founder/CEO of UnboundAnswers) argues most practitioners use AI at the "execution layer" — drafts, summaries, first passes — which is real productivity but not the real return. He calls it a "mode problem," not a tooling problem.
- He grounds it in a peer-reviewed Drexel study (Tim Gorichanaz, 2025 ASIS&T) of 205 real-world ChatGPT use cases across six modes — Writing alone was ~47%, Identifying ~10% (dataset from Reddit, Anglophone-skewed).
- His core line: "You cannot retrieve your way to judgment." Retrieval output now sounds so much like judgment that the gap is nearly invisible — especially to those who haven't built judgment yet.
- Supporting research he cites (Microsoft, Swiss Business School, TestGorilla) links heavy AI reliance to declining critical thinking, strongest in younger, less-experienced practitioners.
You have the AI tools. You're using them daily — for drafts, summaries, first passes that used to take twice as long. That's real productivity. According to Duane Forrester's widely-shared piece, it's also not the return the investment can produce. His framing is sharp: the gap between what you're getting and what's available isn't a tool problem, it's a mode problem. You're operating AI at the execution layer when the leverage lives one layer up.
What the research actually shows
Forrester anchors the argument in a peer-reviewed study by Tim Gorichanaz at Drexel University, presented at the 2025 ASIS&T Annual Meeting, which analyzed 205 real-world ChatGPT use cases and sorted them into six modes: Writing, Deciding, Identifying, Ideating, Talking, and Critiquing. Writing dominated at roughly 47% of observed cases; Identifying was around 10%. (Worth noting the dataset came from Reddit and skews Anglophone.) Separately, he cites that about 63% of organizations using generative AI apply it primarily to create text. The pattern is consistent: we've mostly bolted AI onto the part of the work that was already mechanical.
Execution vs. judgment
The distinction Forrester draws is the whole point. Retrieval and execution — drafting, summarizing, synthesizing what's known — is what AI does well. Judgment is different: knowing which question is actually the right one in this context, recognizing when something that looks correct is wrong for reasons that aren't in any training data, and the accumulated weight of having been wrong in consequential situations and recalibrating. As he puts it, "you cannot retrieve your way to judgment" — you build it through deliberate practice, under real conditions, with skin in the game a model structurally can't have.
Why the gap is dangerous
The risk isn't that AI handles retrieval well. It's that retrieval output now sounds like judgment output — confident, fluent, structured — so the difference has become nearly invisible, especially to people who haven't built enough judgment to feel the seam. A polished AI answer and a hard-won judgment call can read identically on the page. One of them is load-bearing; the other isn't.
The critical-thinking cost
Forrester points to converging research: Microsoft, the Swiss Business School, and TestGorilla have independently documented that heavy AI reliance correlates with declining critical thinking — and the effect is strongest in younger, less-experienced practitioners. That's the compounding danger: the people most likely to lean hardest on execution-layer AI are the ones still building the judgment they're outsourcing.
What to do about it
- Use AI for execution, deliberately reserve judgment for yourself. Let it draft and summarize; you decide which question matters and whether the "correct-looking" answer is actually right here.
- Protect the reps that build judgment. The hard calls — the ones with consequences — are how you develop the thing AI can't. Don't automate the practice away, especially early in a career.
- Differentiate on judgment. As execution-layer output commoditizes, the career and team leverage moves up the stack — to the people who know which question to ask.
Frequently asked questions
What are the "execution layer" and "judgment layer"?
Duane Forrester's terms: the execution layer is AI-friendly work — drafting, summarizing, synthesizing known information. The judgment layer is knowing the right question to ask and recognizing when a correct-looking answer is wrong for the specific context. He argues most people use AI for the former and under-invest in the latter.
What study does the argument rely on?
A Drexel University study by Tim Gorichanaz (2025 ASIS&T) of 205 real-world ChatGPT use cases, which found Writing accounted for ~47% of cases. The dataset was drawn from Reddit and skews Anglophone.
Does using AI hurt critical thinking?
Forrester cites independent research from Microsoft, the Swiss Business School, and TestGorilla linking heavy AI reliance to declining critical thinking, with the strongest effect among younger, less-experienced practitioners.
So should I stop using AI for drafts?
No — execution-layer use is genuinely productive. The point is to keep the judgment work (deciding what matters, validating context) yourself, and to protect the consequential reps that build judgment over time.
Sources
- Search Engine Journal — You're Using AI at the Execution Layer. The Value Is in the Judgment Layer.
- Duane Forrester Decodes (Substack) — original essay
