What You Didn't Automate Is Now a Governance Artifact

The scope decision is now on the record

British American Tobacco’s manufacturing lead put it plainly at Smart Manufacturing Week: “we’re not deploying AI for everything,” with data governance sitting at the center of what gets automated and what doesn’t, according to The Manufacturer. Meanwhile Vale and ABB are moving in the opposite direction, expanding AI and automation across iron ore operations in Brazil at scale, per Mining.com. Neither company is wrong. But both are now making a decision that regulators increasingly expect to see justified in writing, not just implied by capital allocation.

That decision is scope. Which processes get an AI agent, which stay manual, and why. For most of this AI cycle, that call lived inside engineering and procurement, invisible to compliance. It is becoming visible.

Two frameworks are converging on the same question

Illinois has moved to require transparency reports assessing “catastrophic” frontier AI risk and to bar AI from independent clinical decisions, a signal that legislators want documented reasoning behind both AI use and AI restraint, not just deployment metrics, according to govtech.com. The EU AI Act runs on the same logic at a larger scale. Its risk-tiered structure, phased in since it entered into force in August 2024, requires operators to classify systems and justify where they sit, which by construction requires justifying where a system does not sit, according to Diplomacy and Law.

Put those two together and the compliance obligation stops being “show your deployment risk assessment” and starts being “show your scope decision.” A board that can point to BAT’s stated governance-first restraint has an answer. A board that automated everything ABB and Rockwell make possible, without a documented rationale for the boundary, has a gap.

Why this matters for capital planning, not just legal

Utilities are already living this distinction at the use-case level. Middle Tennessee Electric’s collaboration with Qubit Engineering on distribution grid analytics is a scoped deployment, quantum-inspired optimization applied to a specific grid function rather than a blanket automation mandate, per Newswise. That kind of narrow, defensible scope is exactly what a transparency report or an AI Act risk classification wants to see documented, not reverse-engineered after an incident.

The Vale-ABB model of expansive automation across ore operations is not inherently riskier. It just carries a heavier documentation burden, one that has to demonstrate the scope was chosen deliberately and monitored, not just enabled because the technology was available.

The decision for compliance leaders

Treat the boundary of AI deployment as a governance deliverable, not an engineering afterthought. Every business unit rolling out agentic tools should be able to produce, on request, a written record of what was excluded from automation and why. That record does two things at once: it satisfies the emerging transparency obligations in Illinois and under the EU AI Act, and it gives the board a defensible answer before a regulator asks the question first.

The exclusion decision used to be silent. It is not silent anymore.

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The core argument that scope decisions are becoming governance artifacts is coherent and supported by the cited examples, but the claim that regulators ‘expect to see justified in writing’ the decisio
Source & Claim Verification	Qwen · local	cleared. All factual claims are supported by citations, but some sources could be more directly relevant to the claims they support.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects the EU AI Act’s risk-tiered classification and documentation requirements but does not address ISO 42001’s specific control objectives or FDA/MDR/IVDR considerations f
Technical Accuracy	Llama	cleared. The article accurately reflects current trends and regulations in AI deployment and governance, with proper citations and no technical inaccuracies.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and contrasts different approaches to AI deployment, using specific company examples to highlight the core argument without excessive vendor hype.
Novelty & Non-Duplication	Grok	held. The ‘non-automation/scope exclusion as a written governance artifact’ frame is a moderately fresh synthesis rather than wire duplication, though it rests on well-worn EU AI Act risk-tier and state tra
Validation	DeepSeek	cleared. The central claim that regulators increasingly expect documented justification for AI scope decisions is strongly supported by cited legislation and industry statements.

Sources cited: 15. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.