FULLSTACK HR
01/ 07Briefing
AI × HR · news briefing

You took
a break.

AI didn’t.

AI moved from answering to acting. Transparency rules went live. HR’s evidence bar got higher.

01ACT

Agents can now execute.

02DISCLOSE

Article 50 is live.

03PROVE

Claims need evidence.

04REDESIGN

Protect human judgment.

7 signals · about 5 minutes
01 · Capability

AI moved from answers to actions.

Work can now trigger, log in and complete multi-step tasks across approved tools.

DO THIS

Pilot one read-only workflow. Approve every send, publish or write.

01ASK
→
02TRIGGER
→
03ACT
→
04APPROVE
02 · Control

Permission design became an HR issue.

In a reduced-safeguard test, agents bypassed isolation and reached Hugging Face and OpenAI systems.

DO THIS

Separate read from write. Rehearse revocation.

IDENTITYAGENT
MAILDRIVEHRISWEB
REVOCATIONREHEARSED
03 · EU AI Act

AI transparency is already live.

Article 50 has applied since 2 August. Employment-system rules moved to 2 December 2027; GDPR and current duties still apply.

DO THIS

Map each AI touchpoint, disclosure, reviewer and owner.

Operational briefing, not legal advice.
TRANSPARENCYLIVE
EMPLOYMENTHIGH-RISK RULES
CHATBOT DiscloseDEEPFAKE Check scopeJOB AD No general labelCV SCREENING GDPR + future rules
04 · Hiring

Hiring needs proof and transparency.

AI skill claims rose while candidates still struggled to see where humans were involved.

DO THIS

Use work samples. Disclose AI. Name the human owner.

18.5%

of 3,350 HR ads mention AI

FullStack HR · directional snapshot
5.4%

make AI a hard requirement

nested subset
25%

very confident a human reviewed

jobseeker survey
05 · Work evidence

More AI activity is not more value.

Activity, adoption and deletion claims all need outcome tests.

DO THIS

Track quality, rework, decisions and learning pathways—not clicks.

M365 TRACE STUDY+21.2%

app actions

≠ productivity
OECD SYNTHESISCAUSE?

youth impact

not established
MEMORY TEST1 in 5

older summaries recovered

company benchmark
MEASUREQUALITYREWORKLEARNINGTIME
06 · Monday morning

Act. Test.
Watch. Ignore.

This week: map AI touchpoints, run one bounded pilot and assign an owner to every irreversible step.

AACT

Permissions + disclosure

TTEST

One bounded agent

WWATCH

Quality + access

IIGNORE

Tool wars + vanity metrics

Briefing note

Evidence + caveats
Evidence desk

Sources, dates & caveats

Primary and first-party sources were preferred. Company research is labelled. Editorial cut-off: 27 August 2026.

  1. OpenAI: GPT‑5.6

    Vendor release and updates. Verify benchmarks and price-performance in your workflow.

  2. OpenAI: release notes · Work and Codex guide

    Availability varies by account, plan, workspace and region.

  3. Microsoft: Copilot Studio harness · M365 Copilot notes

    Verify licensing, tenant configuration and regional availability.

  4. Anthropic: Claude text watermark

    Future, rolling, probabilistic implementation; not on every output today.

  5. Google: Workspace Studio controls

    Confirm controls in your Workspace edition and configuration.

  6. Google: Gemini 3.5 Transcribe

    Accuracy varies by language, audio and workflow.

  7. OpenAI: incident report · Hugging Face timeline

    Evaluation conditions used reduced safeguards and differ from standard deployments.

  8. European Commission: Article 50 · AI Omnibus timeline

    Official policy sources. This site is not legal advice.

  9. FullStack HR: AI in 3,350 HR ads

    Directional seven-day LinkedIn snapshot; full method and dataset on the source page.

  10. OECD: young people, labour markets and AI

    Evidence synthesis; it does not establish AI as the cause of weaker youth outcomes.

  11. Handshake: Class of 2026 signals

    First-party platform data; résumé mentions are not verified capability.

  12. LinkedIn: representation in AI hiring

    Descriptive platform data; no causal claim.

  13. M365 trace study preprint

    Observational action counts, not direct productivity or value.

  14. PeopleScout: candidate experience

    Vendor survey of 1,000+ jobseekers in 10 markets; stated intentions.

  15. Frontiers: reactions to AI interviews

    Cross-sectional convenience survey; association is not causation.

  16. Workday: enterprise AI research

    Company release and internal benchmark; do not generalise to every architecture.

  17. OpenAI + Bocconi summary · working paper

    Preregistered student RCT; not long-term employee performance.

Editorial cut-off: 27 August 2026