Page 1 · Map
AI Is A Discipline, Not A Shortcut
Set your starting point: where your team stands with AI, the numbers The Engine is meant to move, and the commitment you are making.
A prompt is not a process. A draft is not a deliverable. An AI workflow is real when it has a business goal, a human review, a measured result, and an owner.
The Engine Map
- Pick the right use case
- Load the context
- Prompt and produce
- Review and verify
- Launch and test
- Measure and scale
Maturity before ambition
The self-assessment below is inspired by the five stages of the AI Marketing Canvas (Venkatesan & Lecinski, 2021): Foundation, Experimentation, Expansion, Transformation, Monetization. A stage counts only when every stage before it is solid.
Tool — AI maturity self-assessment (auto-calculating)
Score each statement from 1 (not at all) to 5 (fully true), on evidence you could show a manager.
| Stage | Statement | Score 1–5 |
|---|---|---|
| Foundation | Our customer data is accessible and usable, and we have clear rules on what may go into AI tools. | |
| Experimentation | We run small, deliberate AI pilots on specific marketing tasks, with a named owner. | |
| Expansion | Proven AI workflows are shared and reused across several people, teams, or channels. | |
| Transformation | AI is built into how we plan, produce, review, and measure campaigns. | |
| Monetization | AI capabilities create new value for customers or new revenue for the business. |
Your stage is the last one scored 4 or 5, provided every stage before it is also 4 or 5.
Tool — Baseline metrics & commitment
Baseline — the numbers The Engine is meant to move
| Metric | Today (kickoff) | Final clinic |
|---|---|---|
| Hours per key deliverable | ||
| Brief-to-publish cycle time | ||
| Share of AI drafts that pass review first time | ||
| KPI of my priority use case |
Use it in the work — re-read this commitment at the start of every session.
Tool — Session tracker
| Session | Date | Key action I will take next | Done |
|---|---|---|---|
| Discovery kickoff | |||
| E-Learning | |||
| Live Online Simulation | |||
| Coaching 1 · Scope | |||
| Coaching 2 · Build | |||
| Coaching 3 · Prove | |||
| Team clinics |
Page 2 · Context
Pick The Right Use Case, Load The Context
Not every task deserves AI. Find the ones that do, rank them on evidence, and give the model what it needs to sound like you — without giving it what it must never see.
Tool — AI use-case finder
List the recurring marketing tasks that eat your week. Be as honest about where AI should not help as about where it should.
Practice first: try the Use-Case Finder simulation, then come back with your own tasks.
| Recurring task | Channel | Where AI helps | Where AI should not |
|---|---|---|---|
Tool — Use-case prioritization matrix (auto-calculating)
Score each use case 1–5 per column. Low risk scores high: 5 means the least risk to brand, data, and customers.
| Use case | Business value | Feasibility | Data readiness | Low risk | Total / 20 |
|---|
Use it in the work — bring your top-ranked use case to Coaching 1.
Tool — Brand voice & context pack
Paste this pack at the top of every prompt. Use only facts you have verified.
Tool — Data classification guide
What may go into which tool. These are starting defaults: write your organization's actual rule in the last column.
| Data class | Examples | Approved AI tool | Unapproved AI tool | Our rule |
|---|---|---|---|---|
| Public | Published web copy, press releases | Yes | Yes — still verify the output | |
| Internal | Draft plans, internal briefs, meeting notes | Per policy | Never | |
| Confidential | Pricing, contracts, unreleased products, results | Only if policy allows | Never | |
| Personal information | Names, emails, customer records, CRM exports | Only if policy allows — minimized | Never | |
| Regulated | Health, financial, or other regulated data | Only with legal / compliance sign-off | Never |
When in doubt, leave it out. This guide is not legal advice: adapt it to your organization's policies and jurisdiction.
Page 3 · Prompt
Prompt With Structure, Produce With Intent
Brief the model the way you would brief a new team member: context, objective, style, tone, audience, and the response you expect.
Tool — CO-STAR campaign prompt builder
Inspired by the CO-STAR framework (GovTech Singapore). Fill the six parts; the prompt assembles itself below.
Practice first: judge six prompts in the Prompt Lab simulation, then build your own below.
Tool — Starter prompt library
Replace every [bracket]. Paste your context pack first. Use only your organization's approved AI tools.
Tool — Content variant planner
| Channel | Segment | Variation angle | Call to action |
|---|---|---|---|
Page 4 · Verify
Review Before Anything Ships
A draft is not a deliverable. A named human checks accuracy, brand, bias, privacy, intellectual property, and disclosure — every time.
Human review is the job
AI makes drafting cheap. It does not make publishing safe. The reviewer owns the result as if they had written every word.
Guardrails, inspired by the NIST AI RMF
- Govern — who owns the rules, the tools, and the sign-off.
- Map — where this workflow could cause harm.
- Measure — how you check output quality and risk.
- Manage — what you do when a check fails.
Tool — Customer insight prompts with verification
Verification checklist — before any insight is used
Tool — Human review checklist (auto-calculating)
Accuracy
Brand
Bias
Privacy
Intellectual property
Disclosure
Ready to ship only when all twelve checks are ticked. One open check means hold.
Tool — Review log
| Draft / prompt version | Reviewer | Issue found | Fix applied | Date |
|---|---|---|---|---|
Use it in the work — the issues you log here become the next version of your prompt.
Page 5 · Measure & Scale
Measure What Changed, Then Decide
Name the KPI and the test before launch, read the result against the baseline, and document the workflow so a manager can approve it.
Tool — Test & measurement plan (auto-calculating time saved)
| Input | Value |
|---|---|
| Hours per deliverable — before (baseline) | |
| Hours per deliverable — with the AI workflow | |
| Deliverables per month | |
| KPI — baseline | |
| KPI — result |
Hours saved = (before − with AI) × deliverables per month. Time per deliverable = change vs. before. Review time counts as time spent.
Tool — AI Workflow Card (manager-ready)
30-day action plan
| Window | Focus |
|---|---|
| Days 1–10 | |
| Days 11–20 | |
| Days 21–30 |
Weekly rhythm
Frameworks referenced
Venkatesan, R., & Lecinski, J. (2021). The AI Marketing Canvas: A Five-Stage Road Map to Implementing Artificial Intelligence in Marketing. Stanford Business Books, Stanford University Press. sup.org/books/title?id=32597
GovTech Singapore. CO-STAR framework (Context, Objective, Style, Tone, Audience, Response), featured in its Prompt Engineering Playbook. tech.gov.sg
National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1. doi.org/10.6028/NIST.AI.100-1. Companion: Generative AI Profile, NIST AI 600-1 (July 2024).
Tools are inspired by these frameworks. FieldEdge Academy is not affiliated with, or endorsed by, their authors. Presented by Sol Tanguay (Initiative Marketing).