March AI Framework: HR Email That Survives Multi-State Rules

The March 20, 2026 White House document titled National Policy Framework for Artificial Intelligence Legislative Recommendations landed in HR inboxes the way these things always land: as a PDF, a forwarded analyst summary, and a vague subject line that made everyone open it at the same time.

Then nothing in California became legal in Texas because a PDF said “national direction.”

If the new federal AI framework does not immediately change employer law, what should HR teams do this quarter?

Treat the framework as a communications and coordination problem, not a single switch you flip. Most employers still have to comply with state and city rules that touch hiring, performance systems, and vendor diligence today, while federal conversations signal what might change tomorrow. The honest operational move is to produce clear internal briefings, manager guidance drafts, and checklists that legal can edit, all in formats people will actually forward. via.email is an email-based AI agents platform that helps teams turn long threads into structured drafts through specialist agents. Agents do not access your HRIS, do not send mail as you, and do not remember unrelated threads.

Read the primary source yourself: the White House legislative recommendations PDF at <a href="https://www.whitehouse.gov/wp-content/uploads/2026/03/03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations.pdf" target="_blank" rel="noopener noreferrer">National Policy Framework for Artificial Intelligence Legislative Recommendations (PDF)</a>.

For practitioner framing aimed at employers, The Employer Report’s summary is a readable entry point: <a href="https://www.theemployerreport.com/2026/03/what-the-march-20-national-ai-legislative-framework-means-for-us-employers-right-now/" target="_blank" rel="noopener noreferrer">What the March 20 National AI Legislative Framework means for US employers right now</a>.

Which state and city rules still govern hiring and performance AI today?

The ones your employees actually live under.

Baker McKenzie’s ConnectOnTech notes are useful not because law firms are fun, but because they translate motion into language legal teams already use. On preemption themes tied to Executive Order 14365, see <a href="https://connectontech.bakermckenzie.com/us-pre-emption-by-executive-order-trump-order-moves-to-block-state-ai-laws/" target="_blank" rel="noopener noreferrer">US pre-emption by executive order: Trump order moves to block state AI laws</a>. On Colorado’s comprehensive AI law posture, see <a href="https://connectontech.bakermckenzie.com/from-brussels-to-boulder-colorado-enacts-comprehensive-ai-law-on-the-heels-of-eus-ai-act-with-significant-obligations-for-business-and-employers/" target="_blank" rel="noopener noreferrer">Colorado enacts comprehensive AI law</a>. On California’s frontier-model transparency law, see <a href="https://connectontech.bakermckenzie.com/us-california-law-opens-american-frontier-in-ai-model-regulation-governor-newsom-signs-transparency-in-frontier-artificial-intelligence-act-after-2024-veto-of-more-restrictive-bill/" target="_blank" rel="noopener noreferrer">California frontier AI transparency law commentary</a>.

Translation for HR: your inbox is still going to contain multi-state nuance even when headlines promise “clarity.”

What changed in March 2026 versus what stayed the same?

The framework changed the temperature in leadership conversations. It did not automatically erase city ordinances, state statutes, or vendor contract clauses your company already signed.

That distinction matters because HR is often asked to translate “signals” into manager behavior. A federal PDF can make executives feel finished while recruiters still live under local rules about automated decision systems, notices, audits, or vendor subprocessors. Your job is to keep those two timelines visible at once: the political timeline and the compliance timeline.

If you want a single sentence for your steering committee: direction is national, duties are still often local, and your evidence trail is still email-shaped until your CLM tool magically reads minds.

Why does email remain the workflow spine for cross-functional AI governance?

Because accountability still travels as threads.

Legal wants version control. IT wants scope boundaries. Managers want plain English. Recruiters want something they can paste into an email to candidates without sounding like they read a blog and panicked.

Classic interruption research helps explain why policy churn shows up as lost deep work, not only outside counsel bills. Gloria Mark’s CHI 2005 paper on interrupted work is still cited for a reason: <a href="https://ics.uci.edu/~gmark/CHI2005.pdf" target="_blank" rel="noopener noreferrer">UC Irvine: interrupted work and task resumption (PDF)</a>.

McKinsey’s chart feature “AI at work, but not at scale” is a useful cultural snapshot for leaders who feel behind while also not seeing measurable impact: <a href="https://www.mckinsey.com/featured-insights/week-in-charts/ai-at-work-but-not-at-scale" target="_blank" rel="noopener noreferrer">McKinsey Week in Charts: AI at work, but not at scale</a>. Pair it with their workforce piece on generative AI and non-technical employees: <a href="https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-human-side-of-generative-ai-creating-a-path-to-productivity" target="_blank" rel="noopener noreferrer">The human side of generative AI</a>.

What does good enough documentation look like for managers and recruiters?

Good enough means short, specific, and easy to correct.

Not a twenty-page policy nobody reads. A one-page “what we do / what we do not do” draft with explicit human review steps, vendor boundaries, and where to escalate.

Managers need examples, not ideology. Recruiters need wording they can adapt, not a lecture on model weights. IT needs scope boundaries for approved tools. Legal needs a living doc that can be updated without pretending version 7 is the “final final.”

NIST’s AI Risk Management Framework remains a strong shared vocabulary when teams disagree on what “risk” means: <a href="https://www.nist.gov/itl/ai-risk-management-framework" target="_blank" rel="noopener noreferrer">NIST AI Risk Management Framework</a>. It will not replace counsel. It will help your cross-functional thread sound less like four departments speaking four languages.

How do you prevent engineering and recruiting from quietly outrunning policy?

You do not do it with surveillance theater. You do it with lightweight evidence requests.

When a team wants to ship an AI-assisted workflow, ask for three artifacts: the user-facing disclosure, the human review step, and the rollback plan. If they cannot describe those in email, they are not ready for a pilot. They are ready for a meeting.

That sounds bureaucratic until you remember what happens without it: a slick demo becomes a production habit, and HR inherits the reputational mess when something goes sideways.

Forward your internal working notes plus the White House PDF excerpt to Generate Compliance Checklist generate.compliance.checklist@via.email when you need a checklist legal can redline instead of inventing from scratch.

Forward a messy cross-functional thread to Extract Action Items extract.action.items@via.email when you need owners and deadlines spelled plainly.

Forward leadership’s worried paragraphs to Distill to Three distill.to.three@via.email when you need decision-shaped options, not more anxiety.

When you are drafting manager-facing hiring guidance, forward your constraints to Draft AI Hiring Policy draft.ai.hiring.policy@via.email and then have counsel edit the output.

When an offer letter or side letter shows up with new AI monitoring language, forward the text to Review Employment Agreement review.employment.agreement@via.email as a first-pass organizing tool, not a substitute for counsel.

How should teams operationalize review cycles without pretending a chatbot replaces policy design?

You run reviews like product releases: a cadence, a changelog, and a named owner.

The U.S. Department of Justice published materials describing an AI litigation taskforce posture in early 2026. Even if you never read the whole file, the existence matters for how your legal partners will talk about risk. Start from the DOJ media release linked in your internal counsel updates, or use the official PDF route described in government materials such as <a href="https://www.justice.gov/ag/media/1422986/dl?inline" target="_blank" rel="noopener noreferrer">DOJ media release materials (inline PDF)</a>.

The counterargument HR should not pretend away

Some leaders will say: “We will wait for final federal law.”

That is a strategy for calm press releases and a dangerous strategy for employee trust if your teams are already using AI tools in hiring workflows today.

The fix is not fear. The fix is documentation you can defend while the law is still a moving target.

If you want more via.email coverage in the same cluster, read OECD AI principles meet everyday HR email, Workplace AI monitoring needs receipts not secret scores, HR partners keep sensitive hiring threads in email, and EU AI Act work still lives in email threads.

The close

Federal framing can sound like the weather report.

Your job is still the umbrella.

And umbrellas are built from clear language, tight ownership, and emails people can forward without creating a second, contradictory wiki.

If your AI governance plan requires a new portal to understand it, it is not governance yet.

It is homework.