For much of modern organisational life, a polished document served as evidence of invisible effort.

A tailored cover letter suggested care. A detailed report suggested analysis. A formal legal brief suggested research. A well-written essay suggested understanding. A complete spreadsheet suggested time spent reconciling data and testing assumptions.

Those inferences were never perfect. But they were often useful because polished work was expensive to produce.

AI has changed that equation.

The collapse of a hidden signal

When a credible artifact becomes cheap to generate, its appearance tells us less about the process that created it.

This is not primarily a problem of misconduct. Even honest AI use weakens the old signal. A person may use AI to improve wording around genuine expertise; another may use it to create a plausible artifact without understanding the subject. The final documents can look similar.

The signal has weakened because the link between visible output and the effort behind it has been severed.

Where this matters

The effect appears in several places at once.

Hiring: Applications become cheaper to tailor and submit, increasing volume while reducing the information contained in any single cover letter or CV.

Education: Take-home writing becomes weaker evidence of unaided authorship or understanding.

Professional services: A polished memo or brief may no longer demonstrate that the author independently checked every fact, source, and citation.

Security: Voice, video, and written communication can no longer be treated as strong proof of identity without additional controls.

The common issue is not that all documents are now false. It is that the artifact alone proves less than it used to.

The replacement: prove process, provenance, or presence

When output no longer proves the work, institutions need stronger evidence elsewhere.

There are three broad strategies.

1. Capture the process

Retain drafts, sources, revision history, data lineage, review events, and decision logs. The goal is not surveillance for its own sake. It is to preserve the path from evidence to result.

2. Establish provenance

Use trusted systems, source controls, signatures, credentials, or other metadata to show where information came from and how it changed.

3. Reintroduce presence where needed

Oral defence, live review, dual authorisation, callback procedures, and in-person verification remain valuable when identity or understanding is the thing being tested.

What organisations should not do

The answer is not to ban all AI-generated content or attempt to detect it perfectly. Detection is brittle, and the relevant question is usually not whether AI was used. It is whether the final output is trustworthy for its intended purpose.

A better policy asks:

  • What must be true before this output can be acted on?
  • What evidence should be retained?
  • Which human is accountable?
  • What cannot be inferred from the document alone?

The opportunity

The collapse of proof-of-work creates a new premium on systems that preserve provenance and process.

AIxtract can help organisations retain the evidence behind extracted data, recommendations, and automated actions. It does not restore the old world in which polish implied effort. It supports a better one, in which trust is grounded in traceable evidence rather than surface appearance.

Adapted from the book Cheap Thinking: What AI Makes Abundant, What It Makes Scarce, and Who Captures the Difference by Daoyuan Li, PhD.