AI × HR / Sources and connections

Employees move ahead of the organisation

Date context: Grok run date; not independently verified publication date of cited material

Shadow AI, limited formal HR adoption and new knowledge layers connect to policy, skills and hiring.

Grok · 2026-09-16

AI-generated briefing. Its claims have not been independently fact-checked.

Open the original source

Shadow AI surges, HR AI lags

Exported from Grok. Report text and inline source links preserved; claims have not been independently verified.

48-hour cut (14–16 Sep 2026). The real story is not “HR gets replaced.” It’s: employees already run gen AI, HR is still early-stage, vendors are shipping agents into Copilot/Workday/ServiceNow, and the labor-market data still refuses to match the layoff narrative. For Volvo Group / Trucks that is a People & Culture problem first, a tools problem second.

1. Shadow AI is the operating system now

Deloitte / Ipsos, out today: 25,000 UK workers. Two-thirds have tried ChatGPT, Claude, Gemini or Copilot. Nearly a quarter use them daily. 31% use gen AI without the employer knowing. 17% pay for it themselves. Combined out-of-pocket spend: ~£1bn / $1.4bn a year. Time saved: ~70 minutes/week on search, drafts, summaries. Workers are adopting faster than employers are provisioning.⁠Reuters

Conference Board (14 Sep): 56% of US workers already use gen AI at work; ~9% daily. Only 26% say their company has a gen AI policy (another 23% “in development”). 55% say the output already matches an experienced human.⁠Conference-board

Volvo read: If you wait for a perfect enterprise stack, the shop floor, dealers, product teams and P&C BPs will keep using consumer tools on company data. Policy + approved tools + manager coaching beats another “AI principles” PDF.

2. HR AI adoption is still thin — retention is the actual fire

Gallagher 2026 US Workforce Trends (coverage 13–14 Sep, n≈3,700 employers):

  • 51% limited AI in HR, only 3% extensive.

  • Top uses: TA 49%, L&D 41%.

  • 57% now rank retention as the #1 HR priority; 63% still had ≥10% turnover in 2025.

  • 73% expect more HR AI by 2028, but trust, privacy (72%) and missing ethics reviews (only 45% have done them) lag.⁠Hcamag

So the market is talking agents. Most HR functions are still piloting resume screening.

3. Product / deal cluster: agents go enterprise, not “HR chatbot”

Josh Bersin Company — Galileo Jupiter (15 Sep). Domain HR intelligence now embeds into Microsoft Copilot, Workday, ServiceNow, HiBob, Gloat, plus Claude/Gemini/ChatGPT via MCP. Claim: 200% faster than general LLMs on their HR “golden prompts”; general models hallucinated or were wrong on >50% of those tests. Architecture pitch: grounded research corpus, not open-web slop.⁠Joshbersin

HR Dive (15 Sep) interviewing Bersin + Gartner: agents will eat payroll FAQs and policy lookup. They will not wipe HR. Job postings for HR are flat = rebalancing, not extinction. 250+ HR role types, 400 skills → 100+ possible agents, but “agent sprawl” is the failure mode. Someone still has to instruct, govern and own the architecture. Bersin’s June HR 2030 note (still the frame this week) floated 30–50% smaller HR headcount over years — not next quarter.⁠Hrdive

Pipefy’s prebuilt HR agent gallery (cultural fit, IDP, contracts, comp studies) is last week’s vendor noise, not new science. Same pattern as IBM watsonx HR agents: package the workflow, sell time back. Useful only if it sits inside your approvals and audit trail.

4. Strategy / labor: four futures, rising fear, weak evidence of mass HR cuts

Conference Board — AI and the Labor Force (15 Sep). Four scenarios, not one forecast: gradual augmentation / concentrated gains / massive displacement / uneven disruption. Already: ~41% of US workers and 18% of firms using AI by end-2025. Projection: in three years, 60–70% of cognitive jobs are human+AI, only 15–25% human-only. Employment and wage effects still hard to see in the aggregate. Their line: stop waiting for certainty; build skills, data and shock absorbers now.⁠Prnewswire

Gallup (15 Sep): 27% of US workers now fear tech will make their job obsolete — new high, +7 pts in a year, double 2017. Under 45s: 34%, now tied with “lose benefits” as the top job worry. College and non-college have converged. Traditional layoff/wage fears are flat. Fear is outrunning measured displacement.⁠News.gallup

Gartner recirc this week: if you cut for AI too early, ~30% of those people get rehired by 2029, often more expensive. “Talent remix” beats headcount harvest.⁠Itbrief

That matches what you already know from industrial transformation: cut process before you cut people, or you buy the same skills back at a premium.

5. Hiring is changing what “skill” means

Xobin Human vs AI Skills Report, mid-year 2026 (covered 14 Sep): 72% of skill groups in their 2024 hiring framework are now fully or partly delegable to some model. Analytical/problem-solving almost fully; operations ~28%; leadership/interpersonal still under 50%. Technical assessments: AI-free coding tests dropped from 75–100% of requests (H1 2024) to 25–50% (H1 2026). The live question is no longer “can AI do this task?” It’s what the human must own vs direct vs check.⁠Hcamag

For technician, service, dealer and software roles at Trucks: rewrite scorecards around judgment under incomplete data, safety, customer context — not “can write a clean JD.”

Volvo-specific in this window

Nothing material on Volvo Group / Volvo Trucks + gen AI + HR in the last 48 hours. Closest live signals are older: Group still framing AI as leadership + data + compliance (not a CAIO fetish); Trucks UK work on opening HDV careers to women and ethnic minorities around the zero-emission shift; Italy site work on structured onboarding and school pipelines. The Global Technology Manager, AI Strategy & Governance role in GTT Digital Operations (Gothenburg) is the internal hinge if P&C wants a real partner on agent governance rather than another shadow-IT fight.

What I’d actually do with this week

  • Treat shadow AI as the current state. Map where Volvo people already use consumer models; put approved Copilot/Workday-class agents where the volume is (policy, time-off, JD drafts, first-line TA).

  • Don’t staff a 40-agent zoo. Pick 3–5 workflows with audit trails. Galileo-in-Copilot / Workday is the pattern worth watching, not another standalone chatbot.

  • Retention > automation theater. Gallagher’s 3% “extensive HR AI” is your competitive opening if you can pair agents with manager quality. Gallup’s other finding from 9 Sep: frequent AI users fear displacement more, unless managers are decent.

  • Rewrite hiring for “direct and check,” especially technical and dealer roles.

  • Ignore recycled May op-eds republished today (“last human department”). The data this week is adoption-without-governance, not extinction.

If you want a next pass: I can turn this into a one-pager for Trucks P&C leadership with “do / don’t / watch” against your current stack (Workday vs ServiceNow vs Copilot).

Related themes