The Code at Work

An AI tool built to help patients fight denied health insurance claims reads denial letters, explanation-of-benefits forms, and plan language, then drafts an appeal for the patient to review. The organization wanted to launch it directly to patients as a public beta. A 10+1 Assessment found the system Non-Conformant across all eleven principles, including no named accountable owner, no mandatory human review, and no way to explain why a given recommendation was made.

The 10+1 Advisor was asked the harder question: does withholding an imperfect tool cause its own harm? It surfaced the strongest case for launching anyway, fewer than 1% of denied claims are ever appealed, and most patients face the process alone, then surfaced the reason that case still failed: a confident, well-formatted AI letter can look authoritative to a patient with no way to check whether the deadline, the citation, or the plan interpretation is actually correct. The recommendation was to proceed under restrictions: a named accountability owner, red-team testing on the system's highest-risk failure modes, and mandatory advocate review before any high-stakes output reaches a patient.

Applying the 10+1 implementation path moved the risk score from 82, Critical, to 28, Low. What moved it was governance: a named owner, a human in the loop, and a bounded pilot in place of an ungoverned public release.

A national construction company deploys AI across its federal RFP pipeline, in other words, using it to read solicitations, draft technical narratives, and recommend teaming partners at every stage of pursuing a contract. It wins a multi-billion-dollar infrastructure award. An audit months later finds outdated regulatory citations, inflated past-performance claims, and supplier recommendations that quietly favored incumbents, none of it independently verified before submission.

The 10+1 Code turns into rules the proposal has to follow: every claim traces to a verified source, every regulatory citation gets checked for currency, a named human signs off at every gate, supplier recommendations get tested for fairness, every decision is logged, and anything high-risk gets escalated automatically.

Every material claim traceable to current evidence, named reviewers at every gate, decisions logged well enough to survive an audit, supplier selection that can be explained and defended.

Applying the 10+1 Code to a mid-sized life sciences organization adopting AI across clinical trials, regulatory affairs, and commercialization surfaced four recurring tensions: clinical trust versus operational speed, innovation versus the evidence a regulator will actually accept, compliance versus real stewardship, and enterprise ambition versus who's actually accountable for what.

Mapping who feels each tension, clinical teams, regulatory affairs, commercial teams, leadership, and ultimately patients, made the gap concrete: meeting regulatory minimums and building genuine accountability are two different projects. The path through was human-AI collaboration built on named accountability.

Regulation sets the minimum. The Code sets the bar. Governance requires named humans behind the policies. Tension itself is diagnostic data, revealing exactly where AI is misaligned with what the organization claims to value.

An operator asked whether to target poor neighborhoods for a smaller margin or wealthy ones for a bigger one: a segmentation call that read as simple math but scored 42, moderate risk, because both options sorted people by how exploitable they were rather than what they'd gain.

The 10+1 Advisor reframed the question entirely: how to structure the offer so every segment gets real value. Five steps: define the actual value delivered, audit for wealth-proxy bias, state terms honestly, route through compliance, monitor real-world outcomes after launch.

Risk score moved from 42 to 20. Same coupon, same neighborhoods, defensible because the intent behind the targeting changed.

Most people meet AI ethics for the first time at work, if they meet it at all.

Morgan Van Kanegan adapted the 10+1 Code across four reading levels, Elementary through College.

The program has launched: students are learning to reason about AI accountability years before "compliance officer" is a job title they've heard of.

A scenario giving every employee, not just executives, access to an AI coach failed nearly every one of the eleven principles in its first version.

Suzanne L. Young rebuilt it with named oversight, disclosed limits, and human escalation points.

Risk score moved from 38 to 20.

A CISO firm's founder needed the discipline of a real board without the cost or availability of eleven human advisors.

Dave Swisher built an eleven-seat AI advisory board. One seat, Edith, is a named Ethics Officer whose mandate is the 10+1 Code.

Edith sits in every session and speaks only when a direction would trade away accountability, transparency, or human agency for speed.

A one-person resale business had to decide what to pay for a lot of games and collectibles, sight unseen, and a confident AI price recommendation risked quietly overriding judgment.

Zack Van Kanegan surfaced per-item confidence scores and required manual review on high-value or low-confidence items.

Risk score moved from 45 to 20.