AI Literacy Training for Employees: A Free 60-Minute Team Session on DOL's 5 Areas

A composite scenario assembled from documented cases: At 8:30 AM on a rainy Wednesday in Quincy, Massachusetts, nineteen staff from a municipal housing and workforce office crowd around two folding tables. A radiator clanks beside lukewarm coffee. An HR generalist dims the lights and plays a forty-two-slide compliance video defining neural networks. Under the table, three personal phones glow blue. A case coordinator pastes a landlord dispute email into a free chatbot to soften her tone before 9:00 AM intake. An operations clerk screenshots a budget table into a browser sidebar. At 9:15 AM, everyone signs the attendance sheet. Compliance is checked, yet nobody practiced catching a fake statute, stripping client names, or challenging a voice-cloned payment call.

Survey data show why passive training fails. In KPMG and the University of Melbourne's global study (n > 48,000 respondents across 47 countries, Nov 2024 to Jan 2025), only 47% of surveyed workers had AI training, 66% relied on AI output without evaluating accuracy, 56% made AI-related job mistakes, and 57% of AI-using staff hid their use (KPMG, Apr 2025). Boston Consulting Group (n > 10,600 workers in 11 countries) found only 25% of frontline workers reported adequate guidance, while 54% would use unauthorized tools if restricted (BCG, Jun 26, 2025). In Gallup's 2025 US panel, frequent frontline AI use dipped from 11% in 2023 to 9% in 2025, and only 30% of US workers reported AI guidelines (Gallup, Jun 15, 2025).

You leave with four assets: a clear answer on "four pillars versus five content areas," the mechanism behind why slide lectures fail, a 15-person cost comparison, and a printable facilitator pack. For the ten-pillar map, see our AI Safety at Work Hub.

Are there four pillars or five content areas in the US Department of Labor's AI literacy framework?

The US Department of Labor defines five foundational AI literacy content areas and seven delivery principles in Training and Employment Notice TEN 07-25, issued February 13, 2026 (DOL TEN 07-25; DOL Release). Google's People Also Ask boxes still ask about "four pillars of AI literacy" because older K-12 and university rubrics used four buckets before federal workforce guidance arrived (Google US SERP, Oct 29, 2026).

In TEN 07-25, federal drafters split task exploration from prompt direction so employers, educators, and Workforce Innovation and Opportunity Act (WIOA) boards teach hands-on steering alongside verification (Captain Compliance).

Notice how TEN 07-25 differs from Make America AI-Ready (AI Ready). Built with Arist, AI Ready sends ten-minute daily SMS lessons to any phone, including flip phones, without Wi-Fi or laptops (DOL AI Ready). Solo texts build personal basics well, yet an SMS thread cannot set shared rules inside a twenty-person shop. Coworkers must sit together to agree on which client columns stay out of prompts and which phone number verifies a wire.

TEN 07-25 pairs its five content areas with seven delivery principles: (1) enable experiential learning, (2) embed learning in context, (3) build complementary human skills, (4) address prerequisites such as device or broadband gaps, (5) create pathways for continued learning, (6) prepare enabling roles like managers and counselors, and (7) design for agility. Documenting a role-specific session also supports EU AI Act Article 4, updated by Regulation (EU) 2026/1744 in July 2026 to require measures supporting staff AI literacy (EUR-Lex). For full task scoring beyond an exit check, see Measuring AI Literacy Across DOL's Five Skill Areas.

Why do slide-deck lectures and compliance videos fail to change employee AI habits?

Slide-deck lectures fail to change employee AI behavior because unsafe habits grow from deadline rush, missing approved tools, and overtrust in fluent prose rather than missing definitions. Watching slides on neural weights never slows down a busy clerk at 4:40 PM.

Teaching AI verification by slideshow is like teaching cardiopulmonary resuscitation (CPR) in a theater without a mannequin. Until your palms press two inches deep at 100 beats per minute, your hands lack muscle memory under stress. Adult learners need the same physical proof: typing a clean AI citation into an official database and watching it return zero results.

Where that CPR analogy breaks is sensory warning. A stalled hospital monitor beeps loudly. Language models fail silently. When a chatbot invents a court quote, drops a negation, or logs a client's Social Security number into consumer training data, the screen stays calm and polite. In a 2025 Microsoft Research and Carnegie Mellon study of n = 319 knowledge workers across 936 GenAI examples, higher confidence in AI correlated directly with reduced critical-thinking effort (The Register, Feb 11, 2025).

Four survey findings reinforce why passive seat time fails:

• Minimal training hours: In Slack's Workforce Index (n = 17,372 desk workers), 61% of respondents had under five hours of AI learning, and 48% felt uncomfortable admitting AI use to managers (Salesforce/Slack, 2024).
• Coaching beats solo slides: BCG (n > 10,600 workers) found regular AI use rose sharply when workers received five-plus training hours paired with live, in-person coaching (BCG, Jun 26, 2025).
• Depth and retention caveats: EY's December 2025 survey reported only 12% of surveyed employees received sufficient AI training; workers with 81+ yearly hours reported 14 weekly hours saved against an 8-hour median, though top trainees were 55% more likely to leave in that survey if internal roles stalled (EY, Dec 2025). Both BCG and EY reflect observational surveys rather than randomized trials.
• Workslop cleanup costs: In a BetterUp and Stanford survey (vendor-sponsored; n = 1,150 US desk workers in one release and n = 1,004 in another), 40% of workers received low-effort AI "workslop" in the prior month, averaging nearly two hours and $186 per employee monthly in cleanup (BetterUp, Sep 2025).

How does the ed.cafe 60-minute AI literacy session run minute by minute for a 10-to-30 person team?

The ed.cafe 60-Minute DOL AI Literacy Session splits one hour into five zero-slide paired drills: sorting inputs, catching planted memo flaws, repairing a weak prompt, speaking fraud and wiretap scripts aloud, and scoring a five-item exit check. Print one two-sided drill pack per pair and keep laptops shut for the first twenty-five minutes.

Minutes 0 to 10: Red / Yellow / Green Paste Sort (Areas 1 and 5)

Give pairs nine work snippets: a zoning press release, job-post bullets, a client CSV with home addresses, a scrubbed vendor note, a patient intake line, a config file with an API key, a named disciplinary draft, anonymized FAQ objections, and public OSHA text. Pairs mark each Green (public), Yellow (approved business workspace only, names stripped), or Red (never paste). At Minute 7, show how free consumer accounts train on inputs by default while paid business tiers carry zero-training contracts. Pair with What Not to Paste Into an AI Tool at Work and our 30-Day Shadow AI Amnesty Plan.

Minutes 10 to 25: Live Hallucination & Citation Catch on a Work Memo (Areas 1 and 4)

Give pairs eight minutes and one phone per table to audit a one-page draft memo holding four planted errors: a fake CFR rule, a flipped KPMG stat, a misread search study, and invented dollar savings. Then explain why models guess: standard benchmarks penalize abstaining. On OpenAI's SimpleQA analysis, a model abstaining on 52% of uncertain items posted a 26% error rate, while a low-abstention model hit a 75% error rate (OpenAI, Sep 2025). Across n = 4,326 queries, Google AI Overviews reached a 91% accuracy rate overall yet cited unsupported sources in 56% of accurate answers (TechRepublic, 2026). Follow up with our First Hallucination Lab, How Often Is AI Wrong?, and Check Every Source.

Minutes 25 to 40: Prompt-Fixing & Context Framing Drill (Areas 2 and 3)

Show a seven-word prompt: "Write an email updating the client on delays." Pairs take ten minutes to rewrite it with four rules: sanitized facts only, exact audience and length, an instruction to output [MISSING] instead of inventing dates or fees, and a bulleted claim footer for human review. Reinforce under Area 2 that chatbots fit rough drafting on supplied text, never safety math or solo HR rankings.

Minutes 40 to 52: Live Callback & Meeting-Bot Consent Roleplay (Area 5)

Have Partner A read an urgent voicemail requesting a $28,500 wire routing change; Partner B speaks the directory callback script aloud. Then Partner A brings an uninvited Otter.ai or Fireflies.ai bot into a call; Partner B pauses and removes the bot, protecting staff under all-party wiretap laws like Massachusetts M.G.L. c. 272 s. 99 ($100/day or $1,000 minimum civil damages: MA Law). Share cards from The Callback Rule Script and AI Notetakers in Meetings.

Minutes 52 to 60: Scored 5-Item Exit Ticket & 7-Day Check-In (All 5 Areas)

Score the five-item exit check aloud in ninety seconds, record group accuracy per question, and schedule a fifteen-minute Day-7 Show-and-Fix Huddle before anyone leaves.

How do you adapt the 60-minute workshop for shift workers, clinics, and workforce boards without laptops?

TEN 07-25 Principles 2, 4, and 6 emphasize matching training to physical job conditions rather than assuming solo laptops. Swap the nine paste cards and flawed memo for four settings:

• Retail, hospitality, and warehouse shifts: Run all five blocks unplugged on laminated break-room cards, or split them into four paid fifteen-minute pre-shift huddles. Show why photographing manifests, customer slips, or schedules into phone apps leaks records.
• Clinics and social services: Swap Drill 2 for a fictionalized intake summary holding an invented allergy or unmentioned symptom. Across n = 565 ambient scribe notes audited at NeurIPS 2026, 31.3% of notes held a verified failure (NeurIPS 2026), and UK social workers documented an AI tool inserting "suicidal ideation" never spoken (BASW). Pair with Case Managers Using AI for Case Notes.
• Construction and electrical trades: Focus Drill 3 on keeping chatbots off OSHA 29 CFR 1926 and electrical math. In an April 2026 test of n = 450 AI-written OSHA scripts, 85.3% of scripts ran cleanly yet only 46.7% of the 450 scripts used correct logic (Uddin, Apr 2026). Hand out Never Use a Chatbot for OSHA Safety Math.
• WIOA career counselors: Practice honest resume tailoring, spotting remote-job scams, and defending non-native writers, paired with Preparing Enabling Roles for Counselors.

What does running an internal 60-minute hands-on session cost compared to compliance videoware?

Running the ed.cafe 60-Minute DOL AI Literacy Session internally for 15 staff costs $0 in software fees and $826.50 in loaded payroll time, cutting first-month spend by 41% against compliance videoware while setting local rules no generic video covers.

Assume a 15-person employer with one facilitator at $48.00/hour loaded ($38.40 base plus 25% in employee benefits and payroll taxes), fourteen staff at $38.00/hour loaded, and a fifteen-minute Day-7 huddle:

Local boards can also fund multi-hour incumbent training through WIOA under TEN 07-25 (DOL TEN 07-25). Massachusetts residents get free access through 2027 to the Google AI Professional Certificate via the Massachusetts AI Hub (Mass AI Hub). To pair solo study with live drills, No Worker Left Behind (Boston 501(c)(3), EIN 88-3268312) is forming ed.cafe Boston Circle #1 at ed.cafe/start/, a free 30-chair cohort closing December 1, 2026 (pre-launch as of October 2026; zero cohorts completed yet). Pull extra worksheets in the Staff Room.

Copy-ready facilitator pack: script, three tabletop drills, and 5-item DOL exit check

Print The ed.cafe 60-Minute DOL AI Literacy Facilitator Script, Drill Pack & 5-Item Exit Check on two double-sided pages and pair it with our First-Hour AI Incident Plan.

THE ED.CAFE 60-MINUTE DOL AI LITERACY FACILITATOR SCRIPT, DRILL PACK & EXIT CHECK
Mapped 1:1 to US DOL TEN 07-25 (Feb 13, 2026) | 10-30 Staff | Zero Slides | ed.cafe

SETUP: Approved Tool (Training OFF): [________________] | Callback List: [________________]
60-Min No-Blame Incident Contact: [________________] | Print 1 Drill Pack/pair + 1 Quiz/staff

[MINS 00-10] DRILL 1: RED / YELLOW / GREEN PASTE SORT (DOL Areas 1 & 5)
Say: "No slides today, and nobody is in trouble for past AI use. Free chatbots train
on prompts by default; commercial zero-retention workspaces do not. Take 5 minutes
in pairs to mark these 9 items RED (never paste), YELLOW (approved workspace ONLY,
names stripped), or GREEN (public)."
 1. Published city zoning press release [GREEN]   6. Config file with live API key [RED]
 2. Public job-posting bullets [GREEN]            7. Named staff warning draft [RED]
 3. Client CSV with home addresses [RED]          8. Scrubbed customer FAQ list [YELLOW]
 4. Vendor note with '[Client A]' swap [YELLOW]   9. Public osha.gov text summary [GREEN]
 5. Patient/client intake with ID/diagnosis [RED unless signed HIPAA BAA]

[MINS 10-25] DRILL 2: FOUR-FLAW MEMO CATCH (DOL Areas 1 & 4)
Say: "AI predicts likely words rather than truth. Use 1 phone per pair for 8 mins
to catch all 4 errors in this memo before it goes out."
 FLAWED MEMO: "Per DOL's mandatory Workplace AI Rule, 29 CFR Part 1994 [FLAW 1: Fake
 CFR rule; TEN 07-25 is voluntary], firms with 15+ staff need quarterly AI audits.
 KPMG's 2025 study shows 66% of staff verify every AI output [FLAW 2: Flipped; 66%
 skip evaluation]. Google AI Overviews cite supporting sources 91% of the time [FLAW
 3: 91% overall accuracy hid 56% unsupported citations]. Moving 14 staff to AI saves
 $4,820 monthly [FLAW 4: Unverified made-up figure]."
Rule at Min 20: "Never send a URL, quote, statute, or number you did not open."

[MINS 25-40] DRILL 3A: PROMPT-FIXING & ABSTAIN GUARDRAILS (DOL Areas 2 & 3)
Rewrite "Write an email to the client about our project delay" using 4 guardrails:
 - Audience & Length: "Draft a 110-word plain email from [Role] to [Role]."
 - Sanitized Facts ONLY: "Use ONLY these 3 facts: [1], [2], [3]."
 - Abstain Rule: "Write [MISSING] if any date or fee is absent; do NOT guess."
 - Claim Footer: "List every factual claim in bullets below the draft."
Rule: Never use chatbots for OSHA/engineering safety math or solo HR rankings.

[MINS 40-52] DRILL 3B: SPOKEN CALLBACK & BOT-REMOVAL ROLEPLAY (DOL Area 5)
Speak both scripts aloud with your partner; switch roles at Minute 46:
 - Wire / Login Urgency: "I hear the rush. Per our callback rule, I am hanging up
   now and calling you back on the number in [CALLBACK LIST]."
 - Uninvited Meeting Bot (Otter/Fireflies/Read): "Hold on before we start. Our
   policy requires advance consent from all parties before AI recording, so I am
   removing this bot from the room now."

[MINS 52-60] 5-ITEM SCORED EXIT TICKET (1 Item per DOL TEN 07-25 Area)
Q1 [Area 1 - Principles]: Why do chatbots cite fake cases or URLs?
   A) Bad Wi-Fi  B) Next-word prediction guesses unless told to abstain  C) Watermarks
Q2 [Area 2 - Explore Uses]: Which task fits an approved commercial AI workspace?
   A) OSHA trench math  B) Ranking staff for layoffs  C) Summarizing scrubbed notes
Q3 [Area 3 - Direct Effectively]: Which rule stops invented dates best?
   A) "Be 100% accurate"  B) "Use ONLY pasted facts; write [MISSING] if absent"  C) "Act smart"
Q4 [Area 4 - Evaluate Outputs]: Minimum check on an AI-cited link before sending?
   A) Check domain  B) Open link, read passage, verify numbers & caveats  C) Ask AI twice
Q5 [Area 5 - Use Responsibly]: You pasted client addresses into free ChatGPT 5 mins ago:
   A) Delete & hide  B) Keep row scope, turn off training, kill share link, report <60m  C) Tell AI to forget
KEY: Q1=B, Q2=C, Q3=B, Q4=B, Q5=B. Items <80% anchor our 15-min Day-7 Huddle: [Date/Time].

Where this stops being true

Four boundaries govern when a sixty-minute workshop must yield to coaching or specialists:

• One hour starts habits; coaching sustains them: BCG (n > 10,600) links durable adoption to five-plus hours with live coaching (BCG, Jun 26, 2025). Follow up at Day 7 and Day 30 with our 12-Point Rubric for AI Work Samples.
• Regulated records and safety math need strict gates: HIPAA PHI, 42 CFR Part 2 files, FERPA records, court filings, and OSHA 29 CFR 1926 math require credentialed controls.
• Unionized shops must bargain first: Unilateral AI rules can violate NLRA contracts; an arbitrator ordered POLITICO to dismantle unilateral AI tools (Allwork, Aug 21, 2026).
• TEN 07-25 is voluntary guidance, not legal immunity: Training never replaces wiretap, hiring, or breach compliance reviewed by counsel.

Questions people ask

Key terms

• Abstain instruction: A prompt rule directing AI to write [MISSING] rather than guessing facts absent from pasted text.
• Commercial zero-retention workspace: Paid business terms guaranteeing prompts and uploads do not train public AI models.
• DOL TEN 07-25: DOL's February 13, 2026 advisory defining five AI literacy content areas and seven delivery principles.
• Enabling roles: Supervisors, HR staff, counselors, and instructors whom TEN 07-25 equips to coach safe daily AI habits.
• Make America AI-Ready (AI Ready): DOL's free seven-day SMS course (beta.dol.gov/ai-ready) teaching individual AI basics in daily texts.
• WIOA: The Workforce Innovation and Opportunity Act, whose funds TEN 07-25 encourages using for contextualized AI literacy.
• Workslop: Unverified, low-effort AI output passed to coworkers, shifting cleanup labor onto the recipient.

Sources

1. US DOL ETA, "TEN 07-25: AI Literacy Framework" (Feb 13, 2026): DOL TEN 07-25
2. US DOL Newsroom, "AI Literacy Framework Release" (Feb 13, 2026): DOL Release
3. US DOL, "Make America AI-Ready SMS Course" (2026): DOL AI Ready
4. Captain Compliance, "DOL AI Literacy Framework" (2026): Captain Compliance
5. KPMG & Univ. of Melbourne, "Trust in AI Global Study" (Apr 2025, n > 48,000): KPMG
6. BCG, "AI at Work 2025" (Jun 26, 2025, n > 10,600): BCG
7. Gallup, "AI Use at Work" (Jun 15, 2025): Gallup
8. Salesforce / Slack, "Workforce Index" (2024, n = 17,372): Salesforce
9. Google US Audit, "AI literacy training" SERP (Oct 29, 2026): Google Query
10. Official Journal of the EU, "Reg. (EU) 2026/1744" (Jul 24, 2026): EUR-Lex
11. The Register on Microsoft/CMU, "AI & Critical Thinking" (Feb 11, 2025, n = 319): The Register
12. EY Newsroom, "Talent Gaps & AI Survey" (Dec 2025): EY
13. BetterUp & Stanford, "Hidden Costs of Workslop" (Sep 2025, n = 1,150 / 1,004): BetterUp
14. OpenAI, "Why Language Models Hallucinate" (Sep 2025): OpenAI
15. TechRepublic, "AI Overviews Citations Analysis" (2026, n = 4,326): TechRepublic
16. MA General Court, "M.G.L. c. 272 s. 99": MA Law
17. NeurIPS 2026, "Ambient Clinical Scribes Audit" (n = 565): NeurIPS 2026
18. BASW PSW, "Social Workers & AI Inaccuracies" (2025): BASW
19. S. M. Jamil Uddin, "LLMs on OSHA Math" (Apr 2026, n = 450): arXiv:2604.12311
20. Mass AI Hub, "Free Google Certificates" (Feb 2026): Mass AI Hub
21. Allwork.Space, "Union AI Bargaining" (Aug 21, 2026): Allwork