Module 03: AI at Work

"The best way to predict the future is to create it." — Peter Drucker (often attributed; the spirit matters: intentional adoption beats drift)

Opening scenario: Your industry newsletter says AI will "transform everything." Your team feels behind—but nobody has budget for a moonshot. Meanwhile, a competitor's social posts look sharper, HR mentions "AI-assisted screening," and finance is experimenting with forecasting spreadsheets. What does sensible AI at work actually look like in marketing, HR, finance, legal, and management—without science fiction? This module grounds you in patterns, stories, and activities you can use Monday.

Learning Objectives

By the end of this module, you will be able to:

• Describe typical AI use cases in marketing, HR, finance, legal, and general management—with realistic benefits and limits.
• Tell plain-language transformation stories that separate workflow change from headline hype.
• Identify where human judgment must remain central—and how to supervise AI-assisted work.
• Facilitate a team conversation about pilots using non-technical vocabulary.
• Spot red flags in vendor claims specific to each function.
• Draft a pilot proposal outline (goals, metrics, risks, owners) without writing code.

1. Marketing & Communications: Speed, Personalization, and Brand Risk

Pattern: AI accelerates first drafts, variant generation, SEO scaffolding, and analytics summaries. It does not replace strategy, ethics, or taste.

Real-world examples

• B2B SaaS: Uses assistants to draft three LinkedIn post variants from a technical blog; human unifies voice and checks claims against release notes.
• Local nonprofit: Generates email subject line tests; measures open rates; never invents impact statistics—only comms team adds verified numbers.
• Law firm marketing: AI suggests client alert outlines from a partner memo; partner approves every legal characterization.

Real World Example

A PTA volunteer uses AI to draft event reminders but pulls dates and room numbers only from the official school calendar. One wrong room number erodes trust faster than any clever sentence — system of record wins.

Try This Now

• In ChatGPT or Gemini: "I market a regional bakery. Propose a 4-week content calendar with themes—not specific health claims—and include a column 'human must verify'."

Table: Marketing tasks—assist vs. automate cautiously

Step-by-step walkthrough: campaign brief

1. Paste product facts you know are true.
2. Ask: "Audience: busy parents; channel: Instagram; goal: foot traffic; constraints: no discounts in copy."
3. Request five hook lines + three storyboard beats.
4. Human: pick one line, rewrite in brand voice, photograph real product.

Discussion prompts

• Who owns brand voice when "everyone uses AI"?

2. HR & People Operations: Efficiency vs. Fairness

Pattern: AI can draft job descriptions, interview guides, onboarding checklists, and FAQ bots. High-risk areas include resume scoring and sentiment inference on employees without governance.

Real-world examples

• Mid-size retailer: Uses AI to standardize job postings across stores; HRBPs edit for local labor market language.
• University HR: Chatbot answers benefits questions from an approved knowledge base only—no free-form medical advice.
• Startup: Turns off automated ranking after reading audit guidance; keeps AI for scheduling interview panels.

Try This Now

• "Draft an inclusive job description for an office manager. Avoid biased adjectives; include ADA-friendly language note for humans to verify."

Table: HR AI risk heatmap

Scenario: An AI ranks candidates using past hires who were mostly one demographic. Outcome: entrenching history. Fix: defer ranking; use AI for scheduling and writing support under policy.

Discussion prompts

• What transparency do candidates deserve about AI in hiring?

3. Finance & Operations: Forecasting, Reporting, and Controls

Pattern: AI helps explain variances, draft month-end commentary, and structure analyses from tables you provide. It should not invent ledger entries.

Real-world examples

• CFO office: Assistant turns a CSV summary (sanitized) into narrative for leadership—analyst verifies every number against Excel.
• Franchise owner: Uses AI to compare three lease scenarios described in prose—lawyer checks legal clauses.
• School business office: Drafts board summary bullets from meeting notes—business manager confirms figures.

Try This Now

• Paste a tiny fictional budget table you make up (no real data). Ask: "Explain variances in plain language for trustees."

Table: Finance tasks—safe prompts

Discussion prompts

• Where should AI stop in your approval chains?

4. Legal & Compliance-Adjacent Work (Non-Lawyer Professionals)

Pattern: AI can organize documents, summarize provided text, and draft routine communications. It is not a substitute for licensed legal advice where required.

Real-world examples

• Compliance officer: Uses AI to create training quiz questions from an internal policy PDF—officer checks answers.
• Contract manager: Extracts dates and renewal triggers from pasted clauses—paralegal verifies.
• Teacher (policy context): Drafts field trip permission letter template—admin approves.

Try This Now

• "From this pasted policy excerpt only, list obligations with bullet points. If unclear, say 'unclear from excerpt'." (Paste something short you wrote yourself.)

Step-by-step: contract reading assistant (human-in-the-loop)

1. Break document into sections.
2. Ask: "Summarize section 3 only."
3. Ask: "Flag ambiguous terms—do not interpret law."
4. Lawyer reads source.

Discussion prompts

• How do you label AI-assisted memos so recipients calibrate trust?

5. Management & Strategy: Decisions, Meetings, and Change

Pattern: Managers use AI for agenda design, stakeholder maps, risk brainstorms, and communication—not for substituting accountability.

Real-world examples

• COO: Generates five ways to phrase a reorg rationale; leadership chooses one empathetic version.
• Principal: Plans PLCs with AI-suggested discussion protocols aligned to goals.
• Agency owner: Uses AI to draft a client renewal email with three tier options—owner adjusts pricing.

Try This Now

• "Facilitator guide for a 45-minute retrospective: psychological safety emphasis; include prompts and pitfalls."

Table: Management uses

Transformation story (composite, realistic): A 40-person services firm adopted AI for proposal outlines only. Win rate did not magically double. Proposal time dropped ~25%, allowing senior review earlier. Lesson: measure cycle time and quality, not magic.

Discussion prompts

• What metric would convince you that a pilot worked?

Activities

1. Function swap: Pick a department you do not know; interview one person about two tasks AI could assist—write a one-paragraph summary.
2. Red flag bingo: Collect five vendor phrases (e.g., "fully autonomous"); translate each into plain accountability language.
3. Pilot sketch: One-page pilot with goal, metric, owner, 30-day scope, kill switch.
4. Storytelling: Write a before/after narrative for a single workflow in your job—keep it honest.
5. Ethics pause: List three stakeholders who could be harmed if your pilot succeeds too fast without guardrails.

Your Challenge

Create an "AI at Work" slide-free brief (2 pages max) for your boss:

• Four departments (pick relevant): each gets one approved use, one prohibited use, one metric.
• One transformation story (real or realistic composite) with lessons.
• Change management: training sentence, IT touchpoint, and how staff report mistakes.

Present it verbally in 3 minutes to a peer.

Expanded challenge: step-by-step (do in order)

1. Pick four functions that touch your reality (e.g., marketing, HR, finance, legal-adjacent, ops). For each, write one sentence describing a task that repeats weekly.
2. Label each task: assistive draft, routing/tagging, forecasting/scoring, or none of these—be honest if it is mostly traditional software.
3. For each function, write one approved use that fits your org’s risk tolerance and one prohibited use you would not pilot without legal/IT (e.g., automated eligibility decisions).
4. Choose one metric per function that a skeptic would accept (time to first draft, tickets resolved without escalation, forecast error vs. baseline—not “engagement” alone).
5. Draft one transformation story with a before state, after state, what did not change (important), and lesson in one line.
6. Add change management: one training commitment, one IT/compliance touchpoint, and how someone reports a bad AI output or near-miss without fear.
7. Rehearse a 3-minute spoken version; ask a peer for one hard question and revise one paragraph.

Real-world scenarios (added)

Healthcare (administrative): A clinic manager uses an assistant to turn bullet notes into patient-facing instructions for a class. Win: faster handouts. Risk: outdated prep guidance. Practice: clinical or nursing leadership approves wording; version-date on the PDF.

Education (district office): Communications uses AI for FAQ drafts about enrollment. Win: consistent tone. Risk: wrong deadline dates. Practice: single owner pulls dates from the official calendar only; AI does not invent timelines.

Finance (corporate): FP&A uses AI to narrate provided variance tables for leadership. Win: readable story. Risk: invented “drivers.” Practice: delete any causal sentence not tied to an analyst note.

Legal (operations): A legal ops team uses AI to reformat playbooks and flag ambiguous headings—not to interpret obligations. Win: housekeeping speed. Risk: silent “fixes” that change meaning. Practice: compare redlines; partner sign-off on anything client-facing.

Marketing (B2C): Social team generates hooks for a product launch. Win: volume of ideas. Risk: comparative claims. Practice: legal/comms checklist on superlatives and testimonials.

Comparison table: “Copilot” workflows vs. traditional templates

Discussion Corner

Use these in a team lunch or cohort seminar:

1. Naming: What would happen if we banned the word “AI” for a month and only named tasks and data—what conversations would improve?
2. Fairness: In your function, who is most likely to be skipped when “the AI handled it” (customers, staff, vendors)? What guardrail fixes that?
3. Credit: When a deliverable is better because of an assistant, how should performance and client billing reflect human judgment vs. tooling—without shame either way?
4. Stop rule: What observable signal (errors, complaints, compliance flags) should automatically pause a pilot in your area?

Try This Now (added)

1. Ask your chat tool: “I am a [your role]. List 6 AI-at-work use cases that are ethical and realistic, and 6 that are high-risk in my function—plain English.” Strike any item you cannot verify or govern; keep the rest for a pilot shortlist.
2. Role-play: Paste a fictional vendor blurb: “Our AI fully automates HR decisions.” Draft your email back asking for task, data, accountability, and audit—no jargon.
3. Table flip: Take one approved use from your brief and ask the tool for three ways it could fail operationally; add your fourth failure mode from domain knowledge.

Key Takeaways

• AI at work usually means assistive drafting, routing, and analysis—not autonomous judgment on high-stakes choices.
• Marketing wins on speed; risk lives in claims and voice.
• HR must weigh efficiency against fairness and transparency.
• Finance requires number discipline—models narrate; spreadsheets decide with humans.
• Legal-adjacent tasks need role clarity—education vs. advice.
• Management value is better decisions and communication, not abdication.

Resources

• Partnership on AI — multi-stakeholder perspectives (browse topics relevant to your sector).
• Your industry association ethics guidelines (search: your field + "AI guidelines").
• Internal IT and legal contacts—treat them as partners early.
• Module 07 (strategy) and Module 08 (ethics) for next depth.

Extended Case Study: Municipal Communications

A city communications team uses AI to draft storm-update posts. Guardrails: only approved shelter addresses; two-person review; pre-written templates for common scenarios. Outcome: faster updates; trust preserved because verification is explicit.

Try This Now: Draft a checklist for that team using your chat tool; compare with the bullets above.

Extended Case Study: Professional Services

Consulting firm allows AI on internal research summaries from client-licensed materials only. CRM logs whether deliverables included AI assistance. Client contracts updated for IP and confidentiality.

Cross-Industry Comparison Table

Discussion Prompts (Seminar Set)

• How do you credit human vs. AI labor in performance reviews fairly?
• Should clients be told when AI assisted a deliverable? When and how?
• What is your one-year skill plan if assistants handle first drafts?

Facilitator Guide Snippet (30-Minute Team Meeting)

1. 5 min: What we mean by AI (Module 01 one-liner).
2. 10 min: Each function shares one idea and one fear.
3. 10 min: Pick one pilot; define metric.
4. 5 min: Next steps and owner.

Try This Now: Stakeholder Map

Ask your assistant: "I am introducing AI note summarization to a school faculty. List stakeholders, hopes, fears, and mitigations in a table." Edit the table to match your culture.

Glossary

• Pilot: Time-boxed experiment with success criteria.
• Human-in-the-loop: Human approves or corrects before effects occur.
• Guardrails: Policies, filters, and workflows that reduce harm.
• Shadow IT: Unapproved tools—often a signal of unmet needs and risk.

Quality Gate Before Scaling a Work Use Case

• Legal/compliance reviewed if needed
• Data classification approved
• Training scheduled
• Feedback channel publicized
• Rollback plan exists

When Stories Sound "Too Good"

Ask for operational detail: What tool? Which data? Who approved? What metric moved—and over what period? Silence on those points is a signal.

Reflection

What is one work task you will not give to AI this quarter because your judgment is the product?

Appendix: Sample Pilot Metrics

• Time to first draft of client email
• Edit distance (how much senior staff still changes)
• Error reports per hundred uses
• Employee confidence (short survey)
• Customer satisfaction (if externally visible)

Pick two; ignore the rest until phase two.