Description

This Corpus note captures Abi’s biggest takeaway from Sarah Cridland’s Experimentation Elite keynote: AI does not replace thinking. It exposes it.

Using a murder-story framing, the piece shows how a detective can have evidence, motive and opportunity, yet still reach the wrong conclusion when critical context is missing. That becomes the central AI lesson: AI can only work with the case file it is given. It does not automatically know your intent, history, constraints, hidden evidence or previous decisions. When context is missing, it fills gaps, and fluent guessing is still guessing.

Abi then pushes into the more dangerous problem: not just missing context, but using AI to prove what you already think. The carousel uses Sarah’s “hallucinated gamekeeper” example to show how leading prompts, plausible details and confirmation bias can combine into something that sounds convincing but is not evidence. As the piece puts it: plausible is not proof, and confirmation bias now has autocomplete.

The argument is not anti-AI. AI can make work better through faster drafts, better brainstorming, cleaner summaries, less grunt work and more room for judgement. But it is the scalpel, not the surgeon. The human still needs to define the problem, provide trusted sources, ban gap-filling, ask for uncertainty, verify the evidence and keep training their own brain.

Topics

• Why AI needs context to be useful

• How missing context changes the answer

• Why fluent output can still be wrong

• Hidden knowledge, internal research and “unknown knowns”

• Why experts need to make their thinking visible

• How AI can reinforce confirmation bias

• Why plausible answers are not proof

• AI hallucination and leading prompts

• Experimentation discipline as a model for responsible AI use

• AI as a thinking amplifier, not a replacement for scepticism

• Why the human should remain in control

Best for

UX researchers, experimentation teams, analysts, AI leads, digital strategists, product teams, content strategists

Background

This piece is part of Abi’s Corpus work on AI, context and better digital decision-making. It extends the Corpus focus on evidence and interpretation into day-to-day AI use: the question is not just what AI can produce, but what assumptions, gaps and biases are sitting underneath the output.

It also connects to the wider Corpus view that trust depends on proof, coherence and operational discipline. AI can accelerate useful work, but only when the human provides context, constraints and scepticism. Without that, the machine does not create clarity. It scales whatever thinking was already there, including the bad bits.