AI ACCELERATOR · LEVEL 2CRM JAMAICA GROUP · AI FOR BUSINESS LEADERS · 210 MIN
A practical 3.5-hour workshop for people who decide where AI goes in an organisation. You leave with a scored use-case portfolio, a costed pilot plan, a one-page AI policy and a 90-day roadmap — built during the session, for your own business. Prerequisite: AI Accelerator Level 1 or equivalent hands-on experience with an AI chat tool.
Duration: 210 minutes including a 10-minute break
Delivery: in person, online or hybrid
Devices: any device with an AI chat tool; a spreadsheet helps
Level: intermediate — no coding; strategy and management focus
Group size: 6–20, ideally one or two people per organisation
Practical result: every leader leaves with a scored use-case portfolio, one costed pilot plan, a draft AI policy and a 90-day roadmap.
| Time | Module |
|---|---|
| 00:00–00:25 | 1 · The AI landscape for business |
| 00:25–01:00 | 2 · Finding value: use-case discovery |
| 01:00–01:35 | 3 · The economics of AI |
| 01:35–01:45 | Break |
| 01:45–02:05 | 4 · Governance, risk and policy |
| 02:05–03:10 | 5 · Build lab: your 90-day AI roadmap |
| 03:10–03:30 | 6 · Boardroom pitch and close |
F for full screen, N for speaker notes, T to start the session clock, O for the slide overview, ? for all shortcuts.| → Space Page Down | Next slide |
| ← Page Up | Previous slide |
| Home / End | First / last slide |
| N | Toggle speaker notes |
| T | Start / pause session clock |
| O | Slide overview |
| F | Full screen |
| B | Black screen |
| R | Reveal answers (quiz slides) |
| 1–9 then Enter | Jump to slide number |
| ? / Esc | Show / hide this help |
Tip: open the page twice — one window on the projector in full screen, one on your own screen with notes open. Both can follow the same URL hash (#12) if you refresh.
Preparation, the minute-by-minute run of show, the four live "strategy co-pilot" demonstrations, the roadmap build lab, troubleshooting and the completion standard.
This is a decision-maker's workshop. Participants are owners, managers and team leads who already use an AI chat tool; the goal is not more prompting skill but better decisions about where AI goes, what it costs, what it earns and what rules govern it. Every module ends with something written into the participant's own plan.
| Time | Section | Facilitator approach |
|---|---|---|
| 00:00–00:25 | Module 1 · The AI landscape | 0–5 welcome, baseline question ("what has AI done for your business so far?") · 5–13 what's changed: models, agents, cost curve, what a leader must know · 13–18 three levels of adoption · 18–23 Demo 1: AI maps a value chain · 23–25 debrief. |
| 00:25–01:00 | Module 2 · Finding value | 0–8 where AI pays: the four value patterns · 8–15 the use-case canvas · 15–25 participants list ten use cases for their org (workbook) · 25–32 score with the value / feasibility / risk matrix · 32–35 share top use case in pairs. |
| 01:00–01:35 | Module 3 · The economics | 0–8 what AI costs: seats, usage, people, change · 8–15 the time-study method and the ROI formula · 15–20 Demo 2: AI estimates an ROI and lists its assumptions · 20–30 participants run the ROI calculator on their top use case · 30–35 build vs buy vs wait. |
| 01:35–01:45 | Break | Ten minutes. Return with your top use case and its ROI number. |
| 01:45–02:05 | Module 4 · Governance | 0–6 the risk map: accuracy, confidentiality, bias, dependence, IP · 6–12 the one-page AI policy and Jamaica's Data Protection Act duties · 12–17 Demo 3: AI drafts the policy, facilitator corrects it · 17–20 vendor due-diligence questions (Demo 4). |
| 02:05–03:10 | Module 5 · Build lab | 0–10 choose the three roadmap use cases · 10–30 pilot plan for #1: owner, metric, data, tool, human review, 30-day timeline · 30–45 draft the AI policy with the policy prompt, then edit · 45–55 90-day roadmap and change plan · 55–65 rehearse the three-minute pitch in pairs. |
| 03:10–03:30 | Module 6 · Pitch and close | 0–12 boardroom pitches (3 min each, volunteers or small groups) · 12–17 knowledge check and rubric · 17–20 commitments and feedback. |
Explain: the tools participants used in Level 1 are the surface. Underneath: models that can read documents, images and spreadsheets; "agents" that take multi-step actions; and a price that keeps falling. A leader does not need the mechanics, but must know four things — what the tools can reliably do today, what they still get wrong, what they cost, and that the organisation (not the vendor) is accountable for outcomes.
Three levels of adoption: (1) individual productivity — each person uses AI for drafting and research; (2) team workflows — a repeated process is redesigned around AI with review steps; (3) products and services — AI inside what customers buy. Most SMEs should earn level 1 and 2 before attempting level 3.
Demo 1: map a value chain. Point out what the AI guessed (margins, volumes) versus what the owner actually knows.
Four value patterns: remove waiting (faster responses, quotes, documents); remove repetition (summaries, data entry, first drafts); widen capacity (more customers served per person); improve decisions (analysis of data you already hold). If a use case fits none, it is probably a novelty.
Use-case canvas: who does the task today, how often, how long, what goes wrong, what the AI would produce, who checks it, what data it needs, what could go wrong. Participants write ten in the workbook — speed over polish.
Scoring: value (1–5), feasibility (1–5), risk (1–5, where 5 is the highest risk). Priority = value × feasibility ÷ risk. The workbook computes it. Debrief: the best first pilots are high value, high feasibility, low risk and easy to measure.
Costs: tool subscriptions (per seat per month) or usage-based pricing; the people time to set up, train and review; and change cost — the dip while the team learns. Keep tool prices generic: they change often and vary by plan.
Time study: minutes per task × tasks per week × people × loaded hourly cost × 52 = annual cost of the task today. Savings = that cost × the share of time AI removes (be conservative: 30–50% for drafting tasks) minus review time. ROI = (annual savings − annual cost) ÷ annual cost.
Build vs buy vs wait: buy a tool when the use case is generic; build (configure an assistant, automate a workflow) when the process is specific to you and the data is yours; wait when the risk is high and the value is unclear.
Demo 2: ask AI to estimate an ROI. The lesson is in the assumptions list — every number the AI invented must be replaced with the participant's own.
Risk map: accuracy (hallucinated facts in customer-facing output); confidentiality (staff pasting customer or financial data into public tools); personal data (Jamaica's Data Protection Act 2020 — lawful basis, minimisation, security, data-subject rights; the Office of the Information Commissioner oversees it); bias and fairness in decisions about people; intellectual property and contract terms; over-dependence on one vendor or one champion.
One-page policy: purpose and scope; approved tools; what may and may not be entered (data classification: public, internal, confidential, personal); human review rules by output type; disclosure to customers where relevant; incident reporting; owner and review date. Regulated sectors add their own rules.
Vendor due diligence: where data is stored and for how long, whether it trains the vendor's models, access controls, export and deletion, uptime and support, price change terms.
Each script is also on its own slide with a Copy button. Run the prompt live, read the answer aloud, then show what you would correct. The reference business is fictional.
You are a business analyst. Westmoreland Hardware & Supplies is a fictional hardware store in Savanna-la-Mar, Jamaica with 12 staff, a retail counter, a small delivery van and trade customers (contractors). Map its value chain as a table: Activity · Who does it · Typical weekly time · Where AI could help · What a human must still check. Do not invent figures — mark every estimate as an assumption to confirm.
Estimate the annual return on using an AI assistant to draft quotations for trade customers at a fictional hardware store. Use this method: minutes per quote × quotes per week × staff involved × loaded hourly cost × 52 = annual cost today; savings = annual cost × share of time removed − review time; subtract the annual tool cost. Show the formula with placeholders, then fill it with clearly labelled example assumptions, and finish with a list of every assumption I must replace with my own numbers.
Draft a one-page AI acceptable-use policy for a small Jamaican business with 12 staff. Sections: Purpose and scope · Approved tools · Data you may and may not enter (public / internal / confidential / personal data under Jamaica's Data Protection Act) · Human review rules by type of output · Customer disclosure · Reporting a problem · Owner and review date. Plain language, no legal jargon, under 450 words. Mark any point that should be checked by a lawyer or the business owner.
Facilitator correction to show: the AI will often approve tools by name — replace with "tools approved by [owner]" — and will omit the review date. Add both live.
List the ten questions a small business should ask an AI software vendor before signing, grouped under: data storage and retention · whether our data trains their models · access control and staff accounts · export and deletion · reliability and support · pricing and price changes. One line each, phrased so a non-technical owner can ask them.
From the scored matrix: one quick win (high feasibility, low risk, visible within 30 days), one core process (highest value), one to explore (promising but needs a question answered first). Write them in the roadmap section of the workbook.
Owner · the one metric that proves it worked (minutes saved per task, response time, quotes per day) · baseline today · target in 30 days · data the AI will use and what is excluded · tool (buy / configure / reuse the Level 1 assistant) · human review rule · who is trained · 30-day timeline · stop condition (what would make you halt the pilot).
Participants fill the policy fields in the workbook, press Generate my policy prompt, run it in their AI tool, then edit the draft. The policy is the participant's; the AI only drafts.
Days 1–30 pilot the quick win · 31–60 measure, decide, extend to the core process · 61–90 policy live, team trained, review with the numbers. Name one AI champion, one training moment, one review meeting.
Three minutes: the problem, the use case, the number (ROI), the risk and how it is controlled, the ask (budget, time, decision). Pairs give one piece of feedback each.
| Situation | What to do |
|---|---|
| Participant has no numbers for their business | Use ranges and the workbook defaults; the method matters more than precision. Flag figures to confirm in the pilot plan. |
| Every use case scores high | Force-rank: "if you could only do one in the next 30 days". Re-score risk honestly — customer-facing and personal-data use cases are rarely a 1. |
| Participant wants a customer-facing chatbot first | Allowed as the "explore" item; the quick win should be internal and reviewable. |
| AI policy draft reads like a legal document | Ask the AI for "plain language, under 450 words, for staff who have never read a policy". Remind: this is a working document, reviewed by the owner and, where needed, a lawyer. |
| Group is mixed sectors (retail, professional services, NGO) | Good — pair across sectors; the value patterns are the same. |
| Participants need more time | Protect Steps A–C of the lab; the roadmap and pitch can be completed as homework and sent in for assessment. |
Each participant pitches or submits their roadmap, names one decision they will take within a week, and one number they will measure.
Completion standard: a scored use-case portfolio (at least six use cases), one pilot plan with a metric and a review rule, an ROI estimate with stated assumptions, a draft AI policy, and a 90-day roadmap. Recommended passing score 70/100 including at least 50/80 on the practical (see Assessment).
Type directly into this page — answers save in this browser. The use-case matrix scores itself, the ROI calculator does the arithmetic and the policy generator writes your prompt. Print or download at the end.
List up to ten tasks AI could help with in your organisation. Score each 1–5: Value (5 = large, measurable), Feasibility (5 = data and tools available, easy to check), Risk (5 = customer-facing, personal data, high consequence). Priority = Value × Feasibility ÷ Risk.
| # | Use case (who · task · output) | Value | Feasibility | Risk | Priority |
|---|---|---|---|---|---|
| 1 | — | ||||
| 2 | — | ||||
| 3 | — | ||||
| 4 | — | ||||
| 5 | — | ||||
| 6 | — | ||||
| 7 | — | ||||
| 8 | — | ||||
| 9 | — | ||||
| 10 | — |
Highest priority: —
| Annual cost of the task today | Annual savings after review time | Annual cost of AI (year 1) | Net benefit (year 1) | ROI |
|---|---|---|---|---|
| — | — | — | — | — |
| Risk | How likely (L/M/H) | Control |
|---|---|---|
| Wrong or invented facts reach a customer | ||
| Confidential or personal data entered into a tool | ||
| Unfair or biased output about a person | ||
| Dependence on one vendor or one champion | ||
| Staff resistance or skill gap |
Knowledge check: 10 questions (20 points). Practical roadmap project: 80 points. Recommended passing score 70/100, including at least 50/80 on the practical. This is a proposed course standard, not an external accreditation.
Total: —
Each certificate gets a unique ID and a SHA-256 fingerprint that is written to the Solana blockchain. Anyone can scan the QR code or enter the ID at verify.html to confirm it is authentic. The certificate represents completion of this workshop and is not an externally accredited qualification.
Templates and references for the Level 2 workshop. The Level 1 course and its editable files are at the Level 1 course page.
AI ACCEPTABLE-USE POLICY — [ORGANISATION] Owner: [NAME] Review date: [DATE] 1. PURPOSE AND SCOPE. This policy applies to all staff and contractors using AI tools for [ORGANISATION] work. 2. APPROVED TOOLS. Only tools approved by [OWNER] may be used for work. Personal accounts may not be used for confidential work. 3. DATA. Public and internal information may be entered. NEVER enter: customer records, personal data (names with contact or financial details, TRNs, health information), payroll, contracts, passwords, or anything marked confidential — unless the tool and [OWNER] have approved it in writing. Personal data is handled under the Data Protection Act 2020. 4. HUMAN REVIEW. Anything sent to a customer, any price, quotation, legal, HR or financial text, and any decision about a person must be checked and approved by a named person before use. AI output is a draft. 5. ACCURACY. Verify facts, figures and references before relying on them. Do not present AI output as independently verified unless it has been. 6. DISCLOSURE. Tell customers when they are interacting with an automated system. Do not use AI to impersonate a person. 7. PROBLEMS. Report any data entered by mistake, any harmful or incorrect output that reached a customer, and any suspected breach to [OWNER] the same day. 8. REVIEW. This policy is reviewed every [6/12] months or when tools change.
| Area | Ask |
|---|---|
| Data storage | Where is our data stored, for how long, and who can access it? |
| Training | Is our data used to train your models? Can we opt out in writing? |
| Access control | Can we manage staff accounts, remove leavers and see usage? |
| Export and deletion | Can we export everything and have it deleted when we leave? |
| Reliability | What uptime do you commit to, and what support do we get? |
| Pricing | How is price set, how much notice of changes, and what happens at renewal? |
| Compliance | How do you help us meet the Data Protection Act and our sector rules? |
| Term | Plain-language meaning |
|---|---|
| Agent | An AI system that carries out multi-step tasks (search, draft, update a record) rather than answering a single prompt. Needs clear limits and review. |
| Per-seat vs usage pricing | A fixed monthly fee per user, versus paying for the amount processed. Seats are predictable; usage can be cheaper for light use. |
| Time study | Measuring minutes per task × frequency × people to cost a process before and after AI. |
| Human in the loop | A named person checks and approves AI output before it is used; the organisation stays accountable. |
| Data classification | Labelling information as public, internal, confidential or personal so staff know what may be entered into which tool. |
| Data Protection Act 2020 | Jamaica's law on handling personal data — lawful basis, minimisation, security and individuals' rights — overseen by the Office of the Information Commissioner. |
| Pilot | A time-boxed, measured trial of one use case with a stop condition and a review date. |
| Change cost | The temporary productivity dip and training time while a team adopts a new way of working. |
AI Accelerator Level 2 — AI for Business Leaders is delivered by CRM Jamaica Group. Enquiries: admin@crmjamaica.org · 876-799-7300.
This is general management education. It is not legal advice on the Data Protection Act or any sector regulation. Tool names are the property of their respective owners; features and prices change over time.