AI Is Missing From 9 Out of 10 Codes of Conduct

Fewer than one in ten codes of conduct mentions AI

The report published on 16 September 2026 by advisory firm LRN draws a sharp line: fewer than one in ten organizational codes of conduct explicitly addresses artificial intelligence or technology ethics^[1]. The sample is large: 2,000 full-time employees.

Nearly all respondents say their employer requires them to certify or acknowledge that document. Coverage, then, is total. The content stays silent on precisely the tool that has entered daily work.

Nearly one in five also says their code lacks practical guidance. The number describes a design flaw in how people are supported, rather than a regulatory gap.

Whoever updated those texts covered the legal obligations and left out the operational part.

The timing matters: the report lands in the months when HR platforms are embedding AI features at enterprise scale. Written rules move at a different speed.

The document everyone signs and few consult

Few internal documents match the reach of a code of conduct.

It reaches every new hire, passes through a signature, and ends up filed as proof of compliance. LRN nonetheless finds a declining share of people who have actually used it as a resource. The signature remains universal; consultation falls.

This desk reads the finding plainly: the broadest literacy channel already present in every organization is idling. A text signed by everyone and consulted by few works as a formality, far less as guidance.

The gap between signing and consulting is where the challenge lives. A code that says nothing about AI teaches people one precise thing: that document describes a job other than theirs. The implicit lesson travels faster than any internal campaign.

Choosing to update it costs little and touches everyone. Few change management levers offer this ratio between spend and reach.

What happens to people when guidance is missing

In the absence of a written rule, each person builds their own. The criterion becomes individual; the boundary becomes private.

The cautious type avoids the tool and gives up a real time saving. The bold type uploads sensitive documents into an external system and opens a genuine risk. Both behaviors come from the same void.

The result is patchy adoption, decided by temperament more than by organizational design.

This is the point boards underestimate most often. AI adoption remains an engineered condition rather than an individual choice. People adapt quickly when they are given a clear boundary and a concrete example.

The language of resistance to change explains little in this scenario. Surveys on AI adoption among employees show rising demand for support, aimed squarely at those who lead. The LRN report places that demand inside the document everyone signs.

Trust falling, incidents rising

The context makes the void more costly. An HR Acuity survey, picked up in the same industry analysis, indicates that 55% of workers experienced or witnessed misconduct in 2025, against 41% in 2024. Some 38% report multiple incidents.

LRN adds a second figure: 66% say they can report abuse without fear of retaliation, down from 71% the previous year.

Two indicators moving in opposite directions describe a trust problem. Incidents rise; perceived safety falls.

A reporting hotline covers half the problem. People want to know how an internal investigation actually unfolds: who runs it, what the timeline is, what outcome it produces. LRN flags the absence of that description in most of the codes examined.

A code that explains the investigation process communicates one thing only: here, a report has a follow-up. It is a verifiable promise, with a modest drafting cost.

Adoption and readiness remain two different measures

This is where the leadership gap opens. Adoption counts how many people open a tool. Readiness counts how many use it with judgment inside a redesigned process.

The two populations overlap only partly, and the second remains the smaller one.

An analysis published by HCAMag (the capability doubling companies odds of lasting AI transformation^[2]) identifies a leadership capability that doubles the odds of lasting AI transformation. The message converges with LRN's evidence: the decisive variable sits upstream of the people using the tool.

Meanwhile AI is entering HR platforms at enterprise scale, as Josh Bersin's analysis of the Galileo Jupiter release documents (HR intelligence goes enterprise^[3]). The tools advance; the written guidance lags behind.

The adoption bottleneck, therefore, sits in the readiness of those who lead. Excellent tools with unprepared leadership return less than average tools with AI-literate leadership. The code of conduct is the point where that literacy becomes written and identical for everyone.

What organizations closing the gap actually do

High-functioning organizations treat the code as a working tool, updated as often as the digital tools themselves. Compliance remains a side effect; guidance becomes the purpose.

• Concrete examples of acceptable use for every job family
• Clear rules on which data may leave for external systems
• The role to ask for an opinion in case of doubt
• A description of the internal investigation process, step by step
• An annual review aligned with the tool release cycle

Every item on this list costs drafting hours more than budget. The return arrives quickly: fewer improper uses, more people trying the tool with confidence. A clear boundary widens use instead of narrowing it.

A second element sets these organizations apart: periodic review alongside the people who actually use the tools. The text leaves the legal department and passes through the teams.

Converting internal talent follows the same logic: those who trust the context learn faster.

The code, in this version, stops being a defensive document. It becomes the first AI training module every person receives, at close to zero cost. Few L&D initiatives start with one hundred per cent coverage.

What changes for those who lead

The September 2026 evidence arrives with different implications for every table.

• CEO: bring organizational readiness to the board, not just the adoption rate
• CHRO: turn the code into the first AI literacy module
• CFO: fund a document review with total coverage of the workforce
• Talent & Compensation Committee: monitor trust in reporting and actual use of the code

LRN's number is a boardroom-grade data point. It says the infrastructure supporting people lags behind the technology already adopted. The remedy has a low cost and a total reach.

For a CFO the line item stays marginal: a document review, brief training, use measurement. The return is measured in incidents avoided and working hours recovered.

The design question for anyone leading people is direct: which document, today, tells a person what is acceptable to do with AI in their job? In nine organizations out of ten the answer remains blank. The next review cycle decides how long that blank will last.