L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
AI for Small Business Managers · Module 3 · Lesson 1

Writing AI-Assisted Job Descriptions and Screening Criteria

How the right prompt can surface better candidates before you read a single résumé.

In early 2023, Sweetgreen — the fast-casual salad chain with roughly 6,000 employees — began using AI drafting tools to standardize job postings across its 220-plus locations. Prior to the shift, individual general managers wrote their own descriptions. The results were inconsistent: some postings drew dozens of qualified applicants, others almost none. After piloting AI-assisted drafts that embedded structured screening criteria — specific food-handler certification requirements, availability windows, and bilingual-preferred language — the company reported a measurable reduction in time-to-fill for crew positions. The lesson was not that AI replaced the hiring manager's judgment. It was that structured inputs produce structured outputs, and structured outputs attract candidates who actually fit.

Why Job Descriptions Fail — and What AI Fixes

Most small business job postings fail for the same reasons: they describe the role too vaguely, bury the most important requirements, and use generic language that attracts everyone and filters no one. A posting that says "must be a team player with excellent communication skills" tells a qualified candidate nothing that distinguishes your role from 10,000 others.

AI writing tools — used deliberately — solve this by forcing you to articulate what you actually need. When you prompt an AI with a structured brief (hours, must-have skills, deal-breakers, compensation range, culture context), the model drafts language that is specific enough to self-select candidates. It also catches common legal exposure: vague phrases like "recent grad preferred" or "energetic young professional" can imply age or other protected-class discrimination. AI tools trained on HR compliance literature will flag or avoid these patterns automatically.

The key skill is prompt quality, not AI magic. Garbage in, garbage out still applies. Managers who give the AI a paragraph of context get a paragraph of generic output. Managers who give it a structured brief get a structured, usable draft.

Building a Structured Brief Before You Prompt

Before opening any AI tool, write down answers to six questions: (1) What does this person do in the first 30 days? (2) What prior experience is a hard requirement vs. trainable? (3) What hours, schedule, or physical demands are non-negotiable? (4) What is the compensation range and any non-salary benefits worth mentioning? (5) What has caused the last two people in this role to fail or leave? (6) What one word describes your team culture accurately?

Feed those answers directly into your prompt. A prompt like "Draft a job description for a part-time kitchen prep associate at a 12-seat café. Hard requirements: food handler card, weekend availability 6am–2pm, able to lift 40 lbs. Trainable: knife skills. Pay: $18–20/hr plus tips. Culture: quiet, craft-focused, no drama. Last two hires left because they hated early mornings." produces a draft that filters applicants before you ever read a résumé.

After the AI drafts, your job is to edit for truth. AI will use polished language that may overstate your perks or understate the physical demands. Read every line and ask: Is this accurate? Misleading job descriptions increase early turnover — the new hire arrives to a reality that doesn't match the posting.

AI-Generated Screening Criteria and Legal Guardrails

Beyond the posting itself, you can ask AI to draft knock-out screening questions — short application questions that filter for your hard requirements before a human reviews anything. For example: "Are you available every Saturday and Sunday from 6am to 2pm?" is a simple yes/no that eliminates a large percentage of applicants immediately and fairly.

Legal guardrail: screening questions must be job-related and applied uniformly. You cannot ask about family status, national origin, disability, or religion — even indirectly. Ask AI to review your screening questions for EEOC compliance and it will flag most problematic phrasing. But note that AI is not a lawyer: any final hiring documentation should be reviewed by HR counsel or a qualified professional, especially if your team exceeds 15 employees (the threshold where Title VII federal protections formally apply).

A second area where AI adds value is writing the scoring rubric for résumé review. Ask it to produce a 5-point rubric for the three to five criteria you care about most. This prevents the common bias of giving extra weight to whoever happened to be reviewed first or whoever looks most like the last successful hire.

Critical Limitation

AI-generated job descriptions reflect the patterns in their training data, which includes millions of postings from large corporations. The language may be too formal, too broad, or subtly skewed toward certain demographics. Always read the draft aloud as if you are the candidate. If something sounds exclusionary or inauthentic, rewrite it. The AI draft is a starting point, never a finished product.

Small Business Reality

You are competing with larger employers for the same candidates. An AI-polished, legally clean, specific job description sends a signal that your operation is professional and organized — even if you are a three-person shop. That perception matters to candidates who have options.

Lesson 1 Quiz

3 questions — free, untracked, retake anytime.
What is the most important input that determines the quality of an AI-drafted job description?
✓ Correct. AI output quality tracks directly to prompt quality. A structured brief with hard requirements, schedule, pay, and culture context produces a targeted draft; a vague prompt produces generic output.
✗ The quality of an AI-drafted job description is determined primarily by how specific and structured your prompt is — not the model version, keyword count, or length.
Which of the following is a legally safe AI-generated screening question for a warehouse role?
✓ Correct. Asking about the ability to perform a documented physical job requirement is lawful and job-related. The other options touch protected classes (familial status, national origin, age) and are prohibited under federal EEOC guidelines.
✗ Only option A is job-related and legally safe. Questions about family status, national origin, or age violate EEOC guidelines regardless of how they are phrased.
Why should a manager always edit an AI-generated job description before publishing it?
✓ Correct. AI drafts are trained on large-company postings and may not accurately reflect a small business's real conditions. Inaccurate postings drive early turnover when new hires discover a mismatch.
✗ The real risk is that AI reflects large-corporate norms. Descriptions may overstate benefits, sound too formal, or create false expectations — all of which increase early-turnover risk.

Lab 1 — Job Description Drafter

Build a targeted job posting using structured prompting. Complete 3 exchanges to finish the lab.

Your Task

You are hiring for a specific role in your business. Describe the position to the AI using the six-question framework from Lesson 1 (first-30-days duties, hard requirements, schedule, pay, failure patterns, culture). The AI will draft a job description, help you refine it, and show you how to add screening questions.

If you don't have a current opening, invent a realistic one for your industry — barista, delivery driver, bookkeeper, retail associate, etc. The more specific your brief, the better the output.

Try starting with: "Help me write a job description. The role is [title]. Hard requirements: [list]. Schedule: [days/hours]. Pay: [$X–$Y]. Culture: [one word]. Last hire failed because: [reason]."
AI Lab Assistant HIRING DRAFTER
AI for Small Business Managers · Module 3 · Lesson 2

AI Tools for Résumé Screening and Interview Preparation

Moving faster without moving blind — using AI to read the pile without replacing your judgment.

Between 2018 and 2022, Hilton Hotels reduced its time-to-hire from 42 days to roughly 5 days by deploying HireVue's AI-assisted video interview screening across its hourly roles. Candidates recorded asynchronous video responses to structured questions; an AI model scored them on verbal cues and structured response quality before a human reviewer saw anything. The controversy was real: researchers at the University of Maryland and the AI Now Institute raised concerns that the video-analysis component could encode racial and gender bias through facial-movement scoring. Hilton eventually removed the facial-analysis layer while keeping structured response scoring. The lesson for small business managers is double-sided: AI screening compresses time dramatically, but the specific features you use — and the biases they may carry — matter enormously.

What AI Can and Cannot Do with a Résumé Stack

When you have 40 résumés for one role, AI can help in two distinct ways. First, it can extract and compare structured data: years of relevant experience, listed certifications, gaps in employment, job-title progression. You paste résumé text into a tool and ask it to rate each candidate against your stated criteria on a simple rubric. This is fast, consistent, and removes the fatigue effect — human reviewers score the last résumé they read worse than the first even when quality is equal.

Second, AI can help you identify what questions to ask about each candidate's specific background. If a candidate's résumé shows a 14-month gap, AI can suggest a neutral, job-related way to explore that in an interview. If a candidate's listed experience is from a significantly different industry, AI can suggest bridging questions that test whether their skills transfer.

What AI cannot do: it cannot reliably evaluate motivation, coachability, or cultural fit from text alone. It cannot make a hiring decision for you. And because AI résumé tools are trained on historical hiring data, they can inherit and amplify historical biases — the now-famous case of Amazon's 2018 internal AI recruiting tool, which the company scrapped after discovering it systematically downgraded résumés containing the word "women's" (as in "women's chess club"), illustrates this concretely. Always review your AI screening outputs for demographic patterns.

Using AI to Build Structured Interview Guides

Unstructured interviews — where each interviewer asks different questions based on gut feel — are both legally vulnerable and predictively weak. Research by Schmidt and Hunter (1998, Journal of Personnel Psychology) established that structured interviews predict job performance roughly twice as accurately as unstructured ones. The challenge for small business managers is that building a rigorous structured interview guide takes time most owners don't have.

AI collapses that time. Give your AI tool the job description you drafted in Lesson 1 and ask it to generate: (1) five behavioral questions using the STAR format (Situation, Task, Action, Result); (2) three role-specific technical questions; (3) a one-page scoring rubric for each question. The result is a consistent interview that protects you legally — you can demonstrate that every candidate was assessed on the same criteria — and gives you actual data to compare candidates against each other.

AI can also help you prepare for candidate questions. Ask it: "What questions will a strong candidate ask about this role?" Strong candidates research and push back. Knowing the likely questions lets you prepare honest, specific answers rather than being caught flat-footed on compensation transparency or growth trajectory.

Practical Workflow for a Small Business Manager

A realistic AI-assisted hiring workflow for a manager without an HR department: (1) Use AI to draft the job description and screening questions — 20 minutes. (2) Collect applications; paste résumé text into AI with your rubric and score top 10 candidates — 30 minutes. (3) Use AI to generate your structured interview guide — 15 minutes. (4) Conduct interviews using the guide; take notes on the rubric. (5) After interviews, paste your notes into AI and ask it to summarize where each candidate differed on your criteria — 10 minutes. Total AI-assisted time: roughly 75 minutes of tool use instead of 4–6 hours of unstructured effort.

The human time is interviews and the final decision. Those cannot be delegated to AI — and shouldn't be. Your assessment of whether someone is honest, motivated, and a fit for your specific team is irreplaceable context that no model has.

Amazon's Cautionary Example

In 2018, Reuters reported that Amazon had quietly scrapped an internal AI recruiting tool it had spent years building. The system had taught itself that male candidates were preferable by training on a decade of historical hires — a decade during which the technology industry was predominantly male. The model penalized résumés from all-women's colleges and downgraded profiles mentioning women's organizations. Amazon's corrective edits couldn't fix the underlying bias. Small businesses using off-the-shelf AI screening tools face the same risk in miniature: always audit your outputs for demographic patterns before making decisions.

Lesson 2 Quiz

3 questions — free, untracked, retake anytime.
Why did Amazon abandon its AI recruiting tool in 2018?
✓ Correct. The model learned from a decade of résumés during which Amazon had hired predominantly male engineers. It then penalized résumés referencing women's organizations and all-women's colleges — a textbook example of historical bias amplified by AI.
✗ Amazon scrapped the tool because it learned gender bias from historical hiring patterns, systematically ranking female candidates lower. Speed and copyright were not the issues.
According to Schmidt and Hunter's research, structured interviews predict job performance at approximately what level compared to unstructured interviews?
✓ Correct. Schmidt and Hunter's 1998 meta-analysis found structured interviews roughly doubled predictive validity compared to unstructured interviews — a powerful argument for using AI to build consistent interview guides.
✗ Schmidt and Hunter's research found structured interviews roughly twice as predictive as unstructured ones — a significant advantage that justifies the effort of building a guide even for small businesses.
Which of the following tasks is LEAST appropriate for AI in the hiring process?
✓ Correct. AI can score, suggest, and summarize — but the final hiring decision requires human judgment about motivation, honesty, and team fit that no current AI model can reliably assess. Delegating final decisions to AI also creates legal liability.
✗ The final hiring decision must remain with a human. AI lacks the contextual judgment needed to assess motivation and fit, and delegating hiring decisions to AI creates legal and ethical liability.

Lab 2 — Interview Guide Builder

Generate a structured interview guide and résumé scoring rubric for your role. Complete 3 exchanges to finish.

Your Task

Using the job description from Lab 1 (or a new role), work with the AI to build a structured interview guide. Ask it to create STAR behavioral questions, role-specific technical questions, and a scoring rubric. Then ask it to identify potential bias checkpoints in the questions it generated.

You can also paste in a fictional résumé excerpt and ask the AI to evaluate it against your criteria — then check whether the evaluation feels fair and criteria-based.

Try: "I'm hiring a [role]. Generate 5 STAR behavioral interview questions, 3 technical questions, and a 1–5 scoring rubric for each. Then flag any questions that might create legal or bias risk."
AI Lab Assistant INTERVIEW BUILDER
AI for Small Business Managers · Module 3 · Lesson 3

AI for Onboarding, Scheduling, and Day-to-Day Team Management

Once they're hired, AI keeps them longer — if you use it to reduce friction, not to surveil.

In 2019, McDonald's acquired Dynamic Yield, an Israeli AI personalization company, for approximately $300 million — its largest acquisition in 20 years. While the purchase was primarily aimed at drive-through menu personalization, the underlying workforce scheduling algorithm had already been tested across McDonald's franchise locations to optimize crew scheduling against predicted traffic patterns, local events, and historical order volumes. Franchise operators who adopted the scheduling AI reported labor cost reductions of 2–4% without reducing coverage — meaningful numbers at franchise margins. The mechanism was simple: the AI saw patterns in transaction data that humans missed, and translated those patterns into shift recommendations. Small businesses cannot afford Dynamic Yield, but the logic — schedule to predicted demand, not to habit — is replicable with accessible tools like 7shifts, Homebase, or even a well-prompted general AI assistant.

AI-Assisted Onboarding: Reducing the First-90-Days Cliff

Studies by the Brandon Hall Group found that strong onboarding programs improve new-hire retention by 82% and productivity by over 70%. Yet most small businesses onboard by handing a new employee a binder and hoping they ask questions. AI can radically improve this with almost no capital investment.

The most accessible approach: use AI to convert your existing standard operating procedures, training notes, and policy documents into a structured onboarding checklist with day-by-day tasks for the first two weeks. If you don't have written SOPs, dictate your process to an AI verbally (via a transcription tool) and ask it to convert your explanation into a step-by-step guide a new employee could follow independently.

More advanced small businesses are using AI chatbots embedded in tools like Notion, Slack, or Guru to answer new-hire questions on demand. Instead of a new employee interrupting a manager 15 times on their first week, they ask the bot: "What's the opening checklist?" or "How do I process a return?" The manager reviews the AI's answer log weekly and corrects any errors. This system improves with every new-hire cycle as edge-case questions get captured and answered.

Scheduling Optimization Without Expensive Software

Purpose-built scheduling tools like Homebase (free tier available), 7shifts (restaurant-focused), and When I Work integrate light AI to predict staffing needs against historical traffic. For a small business running on tight margins, even a 1–2% labor cost reduction can be meaningful. These tools pull your POS transaction data, identify your peak and trough periods with more precision than gut feel, and suggest shifts accordingly.

If you can't invest in purpose-built tools, a general AI assistant can still help. Export your weekly sales data to a spreadsheet, paste it into an AI chat, and ask: "Based on this data, what are my three busiest 2-hour windows each day of the week? What shift pattern would minimize my payroll while covering those windows with at least two staff?" The AI cannot access your POS directly, but it can analyze pasted data and produce a recommendation in minutes.

The most common mistake: scheduling based on last week rather than predicted demand. Holidays, local events, weather, and seasonal patterns all affect your traffic in predictable ways. AI tools that have access to calendar and weather data can incorporate those variables automatically. Most managers don't because building the mental model manually takes time they don't have — AI collapses that analysis.

Day-to-Day Team Communication and Performance Tracking

AI writing assistants help managers communicate more clearly and more consistently with their teams. Common use cases: drafting shift change notices, writing policy update announcements that are firm without being harsh, responding to team member complaints in writing in a way that is documented and legally appropriate, and composing performance feedback that is specific and constructive rather than vague and personal.

On performance tracking: AI tools embedded in platforms like 15Five, Lattice (enterprise), or simpler tools like Notion AI allow managers to log brief weekly notes about each team member's performance and get AI-generated summaries that identify patterns. "Over the last six weeks, three of Marcus's shifts included a late arrival" is a fact pattern that's easy to miss week-to-week but obvious in aggregate. These records are also essential legal documentation if a termination is ever challenged.

Surveillance vs. Support

There is a meaningful ethical line between using AI to support your team (better scheduling, clearer communication, faster onboarding) and using it to surveil them (keystroke monitoring, GPS tracking during off-hours, productivity scoring by algorithm). The former builds trust; the latter destroys it. Employee monitoring laws also vary significantly by state — several, including California, New York, and Connecticut, have explicit transparency requirements. Before deploying any monitoring AI, disclose it clearly in writing and consult applicable law.

Practical Priority

If you can only pick one AI application for team management this quarter, pick onboarding documentation. It has the highest ROI, requires no ongoing subscription, and directly addresses the single largest cause of small business turnover: new employees who don't know what they're supposed to do.

Lesson 3 Quiz

3 questions — free, untracked, retake anytime.
According to Brandon Hall Group research, strong onboarding programs improve new-hire retention by approximately how much?
✓ Correct. Brandon Hall Group found that companies with strong onboarding programs improved new-hire retention by 82% and productivity by over 70% — making onboarding documentation one of the highest-ROI uses of AI for small businesses.
✗ Brandon Hall Group research found strong onboarding programs improve retention by 82%. This is why AI-assisted onboarding documentation is one of the highest-ROI applications for small business managers.
McDonald's acquisition of Dynamic Yield in 2019 demonstrated which AI workforce management principle?
✓ Correct. The key insight from McDonald's scheduling AI was that pattern-matching against transaction data — rather than scheduling by habit — yielded 2–4% labor cost reductions. The principle, not the price tag, is what small businesses should take from it.
✗ The McDonald's/Dynamic Yield case demonstrated that AI scheduling based on predicted demand patterns — rather than habit — can reduce labor costs 2–4% without hurting coverage. The principle scales down to small businesses using accessible tools.
What is the primary ethical distinction between AI for team support vs. AI for surveillance?
✓ Correct. The ethical line is between tools that help employees succeed (scheduling clarity, faster onboarding, clearer communication) vs. tools that monitor behavior to enforce compliance. Legal requirements for disclosure also vary by state, with California, New York, and Connecticut having explicit rules.
✗ The distinction is functional and relational: support AI helps employees do their jobs better and builds trust; surveillance AI monitors behavior and erodes it — and carries legal risk in states with explicit monitoring disclosure laws.

Lab 3 — Onboarding Doc & Schedule Optimizer

Convert your procedures into a structured onboarding plan. Complete 3 exchanges to finish.

Your Task

Choose one of two exercises. Option A: Describe a role and its daily/weekly tasks to the AI and ask it to build a structured 2-week onboarding checklist with day-by-day milestones. Option B: Paste fictional weekly sales data (e.g., hourly transaction counts by day) and ask the AI to identify peak windows and suggest an optimal shift schedule.

For either option, push back on the AI's first output — ask it to add a section, change the format, or explain a recommendation. Practicing iterative refinement is the core skill.

Option A: "I'm onboarding a new [role]. Their daily tasks are: [list]. Build a 2-week onboarding plan with daily goals, who they shadow, and checkpoints." · Option B: "Here's my weekly sales data: [paste]. Identify my 3 busiest 2-hour windows per day and suggest a minimum-coverage shift schedule."
AI Lab Assistant ONBOARDING & SCHEDULING
AI for Small Business Managers · Module 3 · Lesson 4

Fairness, Compliance, and Building an Ethical AI Hiring Policy

The legal and ethical framework that keeps AI a tool — not a liability.

On July 5, 2023, New York City became the first jurisdiction in the United States to enforce a law specifically regulating the use of automated employment decision tools (AEDTs). Local Law 144 requires any employer or employment agency using AI or algorithmic tools to make or substantially assist in hiring or promotion decisions in NYC to: conduct an annual bias audit by an independent third party, publish the audit results publicly, and notify candidates that an AEDT is being used. The law was years in the making and imperfect in its first year of enforcement — the Department of Consumer and Worker Protection struggled with audit-standard ambiguities. But the signal it sent was unambiguous: government regulation of AI in hiring is no longer hypothetical. Illinois, Maryland, and California have passed or proposed similar legislation. The window for self-regulation is narrowing, and small businesses that build ethical AI hiring practices now will have a compliance head-start when federal rules eventually arrive.

The Legal Landscape: What Governs AI Hiring Now

At the federal level, the existing framework governing AI hiring is an application of established employment law, not AI-specific statute. Title VII of the Civil Rights Act prohibits employment discrimination based on race, color, religion, sex, or national origin — and applies to algorithmic tools if their outputs produce disparate impact on protected classes, regardless of intent. The Age Discrimination in Employment Act (ADEA), the Americans with Disabilities Act (ADA), and the Genetic Information Nondiscrimination Act (GINA) each impose additional constraints that can be triggered by AI screening tools.

The EEOC issued guidance on AI and algorithmic fairness in May 2023, clarifying that employers are responsible for the discriminatory impact of AI tools they use — even tools from third-party vendors. The critical implication: you cannot outsource legal liability by purchasing a vendor's tool. If the tool produces discriminatory outcomes, the employer bears responsibility. Small business managers must ask vendors directly whether their tools have been audited for disparate impact and request documentation.

State-level law is moving faster than federal. Beyond New York City's Local Law 144, Illinois passed the Artificial Intelligence Video Interview Act in 2020 (requiring consent and disclosure for AI-analyzed video interviews), and Maryland requires disclosure for facial analysis tools. Managers hiring across state lines face a patchwork of requirements that will only grow more complex.

Building a Practical AI Hiring Policy for a Small Business

You don't need a legal team to build a basic AI hiring policy. You need answers to four questions, documented in writing: (1) What AI tools are we using in hiring and at what stage? (2) Who reviews AI outputs before they affect a candidate decision? (3) How do we notify candidates that AI tools are used? (4) How do we audit our AI-assisted hiring outcomes for demographic fairness?

On notification: a simple sentence in your application confirms legal good faith in most jurisdictions: "We use AI-assisted tools to help draft job descriptions and organize candidate information. All hiring decisions are made by a human manager." This is honest, not alarming to candidates, and documents your intent.

On auditing: if you hire more than a handful of people per year, keep a simple spreadsheet of every candidate who applied, their demographic category (if voluntarily disclosed), whether they were screened in or out at each stage, and the stated reason. Reviewing this annually for patterns — even informally — is the minimum viable bias audit for a small business. If you consistently screen out one demographic group at the résumé stage, that's worth investigating before you receive a complaint.

Using AI to Draft Your Own Ethical AI Hiring Policy

The practical skill that closes this module: using AI to draft the very policy that governs AI use. This is not circular — it's efficient. Ask your AI tool to draft a one-page AI hiring policy for a small business in your industry and state, covering disclosure, human oversight requirements, data retention, and annual review procedures. The AI will produce a solid first draft based on known compliance frameworks. Your job is to review it with a lawyer (at minimum, a one-hour consult) and adapt it to your actual tools and workflow.

Key elements to include: list of specific tools used and their purposes; human review checkpoint for every AI-assisted screen; candidate notification language; data retention period for application materials (typically recommended at 1 year under EEOC guidelines); annual review date; and name of the person responsible for policy compliance (usually the owner or general manager in a small business).

Final principle: the goal of an ethical AI hiring policy is not compliance theater — it is building a hiring process that consistently finds the best candidates regardless of who they look like on paper. AI tools, used with structured criteria and human oversight, genuinely help you do that. The policy formalizes the intention.

NYC Local Law 144 — What It Means for You

Even if you're not in New York City, Local Law 144's requirements reveal what best practice looks like: independent bias audits, public disclosure of audit results, and candidate notification. Using these as your voluntary standard — before your jurisdiction requires it — demonstrates good faith and reduces legal risk. The audit requirement alone (annual disparate-impact analysis by candidate demographic) is achievable for most small businesses with a spreadsheet and a few hours per year.

Module Takeaway

AI-assisted hiring and team management is not about automation for its own sake. It is about making structured, consistent, documented decisions that are fairer than gut feel and faster than manual process. The small business manager who combines AI efficiency with genuine human judgment — and builds the policy infrastructure to prove it — has a durable competitive advantage in both hiring quality and legal defensibility.

Lesson 4 Quiz

3 questions — free, untracked, retake anytime.
New York City's Local Law 144 (2023) requires employers using AI hiring tools to do which of the following?
✓ Correct. Local Law 144 requires: (1) annual bias audit by independent third party, (2) public disclosure of audit results, and (3) candidate notification that AI tools are being used. It does not ban AI — it structures its use.
✗ Local Law 144 requires annual independent bias audits, public disclosure of results, and candidate notification. It is the first U.S. enforcement of AI-specific hiring regulation and sets a template others states are following.
Under current EEOC guidance (2023), who bears legal responsibility if a third-party AI hiring tool produces discriminatory outcomes?
✓ Correct. The EEOC's 2023 guidance is unambiguous: employers are responsible for discriminatory impact produced by AI tools they use, even if those tools come from third-party vendors. You cannot outsource legal liability by purchasing software.
✗ The EEOC's 2023 guidance places responsibility squarely on the employer using the tool. Vendor liability may also exist, but the employer cannot escape responsibility for discriminatory AI outcomes by pointing to a vendor.
What is the EEOC's general recommendation for how long employers should retain application materials?
✓ Correct. EEOC guidelines generally recommend retaining application materials for approximately one year after the hiring decision. This enables investigation of any discrimination complaint filed within the statutory window.
✗ EEOC guidelines recommend retaining application records for approximately one year — long enough to cover a discrimination complaint investigation period. Shorter retention creates compliance risk.

Lab 4 — AI Hiring Policy Drafter

Build a compliant, practical AI hiring policy for your business. Complete 3 exchanges to finish.

Your Task

Work with the AI to draft a one-page AI hiring policy for your business. Tell it your state, your industry, your typical team size, and which AI tools you use (or plan to use) in hiring. The AI will generate a draft covering disclosure, human oversight, data retention, and annual review. Your job is to push back — ask it to strengthen a section, simplify the language, or add a specific compliance element you've learned about in this module.

This is a real deliverable: the output of this lab, reviewed by a lawyer, could become your actual AI hiring policy.

Try: "Draft a one-page AI hiring policy for a [industry] business in [state] with [X] employees. We use AI to draft job descriptions and score résumés. Include: disclosure language, human oversight requirements, data retention period, and annual review process."
AI Lab Assistant POLICY DRAFTER

Module 3 Test

15 questions — score 80% or above to pass.
1. What is the primary benefit of using a structured brief before prompting AI to write a job description?
✓ Correct. A structured brief containing hard requirements, schedule, pay, and culture context produces a targeted draft that filters candidates before a human reads a single résumé.
✗ The primary benefit is specificity: a detailed brief produces output that self-selects qualified candidates and filters those who won't fit — before any human review time is spent.
2. Which of these phrases in a job description could signal illegal age discrimination?
✓ Correct. "Young professional" implies age preference and can constitute age discrimination under the ADEA, which protects workers 40 and older. AI tools trained on HR compliance literature will flag this pattern.
✗ "Energetic young professional" implies an age preference and can constitute age discrimination under the ADEA. AI compliance tools flag this — one of the clearest practical benefits of AI-assisted job description drafting.
3. Sweetgreen's 2023 AI-assisted job posting initiative demonstrated what key operational lesson?
✓ Correct. Sweetgreen's shift from inconsistent manager-written postings to AI-assisted standardized drafts with embedded screening criteria reduced time-to-fill for crew positions — because specific postings attract specific, qualified candidates.
✗ Sweetgreen's case showed that standardized, specific AI-assisted postings attracted more fitting candidates and reduced time-to-fill. Specificity drives applicant self-selection.
4. Amazon scrapped its internal AI recruiting tool in 2018 because it:
✓ Correct. Amazon's tool trained on a decade of hiring data from a male-dominated industry and learned to penalize résumés referencing women's organizations and all-women's colleges — a canonical example of historical bias amplified by AI.
✗ Amazon's tool amplified historical gender bias from its training data. It penalized résumés with "women's" in them and downgraded all-women's college graduates — a bias that corrective edits couldn't fix.
5. What format does the lesson recommend for AI-generated behavioral interview questions?
✓ Correct. STAR (Situation, Task, Action, Result) is the industry-standard format for behavioral questions. AI can generate STAR-formatted questions for any role quickly, and they are legally defensible because they focus on demonstrated past behavior.
✗ The STAR format (Situation, Task, Action, Result) is the recommended structure. It focuses on past behavior — a more reliable predictor of future performance than hypothetical questions — and AI generates STAR questions effectively with a clear role brief.
6. According to Schmidt and Hunter's research, how do structured interviews compare to unstructured interviews in predicting job performance?
✓ Correct. Schmidt and Hunter's 1998 meta-analysis established that structured interviews roughly double the predictive validity of unstructured ones — a foundational justification for investing time in AI-generated interview guides.
✗ Schmidt and Hunter found structured interviews roughly twice as predictive as unstructured ones. This research is the foundational argument for using AI to build consistent interview guides even when time is limited.
7. Which AI-assisted scheduling tool approach is accessible to a small business without a POS integration?
✓ Correct. Without expensive software integrations, a small business manager can export sales data, paste it into a general AI assistant, and get a peak-window analysis and shift recommendation in minutes — applying the core McDonald's scheduling logic at zero additional cost.
✗ The practical no-cost approach is pasting exported sales data directly into an AI chat and asking for peak-window identification and shift recommendations. Dynamic Yield is enterprise-scale and not accessible to small businesses.
8. Brandon Hall Group research found that strong onboarding programs improve new-hire productivity by over:
✓ Correct. Brandon Hall Group found strong onboarding programs improve productivity by over 70% and retention by 82%. These numbers make AI-assisted onboarding documentation the highest-ROI team management application for most small businesses.
✗ Brandon Hall Group found productivity improvements over 70% and retention improvements of 82% from strong onboarding — making onboarding documentation the highest-ROI AI application for most small businesses.
9. Illinois passed the Artificial Intelligence Video Interview Act in 2020, which requires employers to:
✓ Correct. Illinois's AI Video Interview Act requires employer disclosure and candidate consent before AI tools analyze video interview content. It was among the first state-level AI hiring regulations and preceded NYC's Local Law 144.
✗ Illinois requires consent and disclosure — not a ban. Employers must tell candidates that AI will analyze their video responses and obtain consent before proceeding. This is the disclosure-based regulatory model most states are now adopting.
10. Under current EEOC guidance, Title VII's disparate impact provisions apply to AI hiring tools:
✓ Correct. Disparate impact law does not require discriminatory intent — only discriminatory effect. If an AI tool produces outcomes that disproportionately screen out a protected class, the employer faces liability even if the discrimination was unintentional.
✗ Disparate impact does not require intent — only effect. If AI tools produce outcomes that disproportionately disadvantage a protected class, liability attaches regardless of the employer's intentions or the tool's design.
11. What is the minimum viable annual bias audit recommended for a small business using AI hiring tools?
✓ Correct. For a small business, the practical minimum is a spreadsheet tracking every candidate, their voluntary demographic information, their stage-by-stage outcomes, and the stated reasons — reviewed annually for patterns that might indicate systematic bias.
✗ The practical minimum is a candidate tracking spreadsheet reviewed annually. You don't need an outside firm — you need a consistent record and the discipline to review it for demographic patterns once a year.
12. Hilton Hotels reduced time-to-hire from 42 days to approximately 5 days using AI screening but later removed which controversial feature?
✓ Correct. Hilton removed the facial-analysis scoring layer from HireVue after researchers raised concerns that analyzing facial movements could encode racial and gender bias. They kept structured response scoring, which is more defensible. This illustrates that not all AI features within a tool are equally appropriate to use.
✗ Hilton removed the facial-movement analysis layer after bias concerns were raised by researchers at the University of Maryland and the AI Now Institute. This case shows that individual features within AI tools carry different risk levels.
13. What should a compliant AI hiring policy always include regarding human oversight?
✓ Correct. Every AI hiring policy should specify that a human reviews AI outputs before they affect any candidate decision. This is both an ethical requirement and the legal mechanism by which you maintain employer accountability for outcomes.
✗ A defined human review checkpoint at every AI-assisted stage is the core structural requirement. Without it, you cannot claim meaningful oversight — and regulators (and courts) will agree.
14. Why is "scheduling based on last week's staffing" described as the most common small business scheduling mistake?
✓ Correct. Last-week scheduling ignores predictable demand variables that AI can model: holidays, local events, seasonal trends, and day-of-week patterns. These are consistent, foreseeable factors that humans skip because the analysis takes time — time AI collapses to seconds.
✗ The mistake is ignoring predictable demand variables. Holidays, local events, weather patterns, and seasonal cycles affect traffic in foreseeable ways. AI scheduling tools incorporate these variables; habit-based scheduling ignores them.
15. What candidate notification language does the module recommend as a legally good-faith standard for small businesses using AI in hiring?
✓ Correct. A straightforward one-sentence disclosure — "We use AI-assisted tools to help draft job descriptions and organize candidate information. All hiring decisions are made by a human manager." — documents good faith, is not alarming to candidates, and satisfies the transparency spirit of current and emerging regulations.
✗ The recommended approach is a simple, honest one-sentence disclosure in the application. It documents intent, is not alarming to candidates, and satisfies the transparency requirements of current and emerging AI hiring regulations.