AI Workflow Canvas

The 5-Column Canvas

The AI Workflow Canvas helps you systematically identify, evaluate, and prioritise AI opportunities across any business process. Use it in workshops, strategy sessions, or individual planning.

Column	Question It Answers	How to Fill It
1. Workflow Step	What are the discrete steps in this process?	Break the process into 5–15 individual steps. Be specific — "review data" is too vague; "compare this month's figures against last month's targets" is actionable.
2. Current State	How is this step done today?	Note the tool used, time taken, who does it, and pain points. Capture the reality, not the ideal.
3. AI Opportunity Type	What kind of AI enhancement fits?	Choose from: Automate, Augment, Accelerate, or Analyse. Most steps are Augment (human + AI), not Automate (AI alone).
4. Tools & Prompts	What AI tools and prompts would you use?	Specify the tool (ChatGPT, Claude, Copilot, etc.) and draft a starter prompt for each step.
5. Priority Score	Should we tackle this first?	Score 1–5 on: Time Saved × Frequency × Ease of Implementation. Higher = do first.

AI Opportunity Types Explained

Automate: AI handles the step end-to-end with minimal human input. Best for repetitive, rule-based tasks. Example: auto-categorising support tickets.
Augment: Human and AI collaborate. AI drafts, human refines. Best for creative and strategic tasks. Example: AI drafts a report, you edit and add insight.
Accelerate: AI speeds up a step you'd do anyway. Best for research and data gathering. Example: AI summarises 20 articles in 5 minutes instead of 2 hours.
Analyse: AI finds patterns or insights in data. Best for large datasets. Example: AI identifies trends in 12 months of customer feedback.

Worked Example: Monthly Reporting Workflow

Here's how a finance team mapped their monthly reporting process using the canvas:

Workflow Step	Current State	AI Opportunity	Tools & Prompts	Priority
Gather data from 4 systems	Manual export, 2 hrs, error-prone	Automate	Power Automate + Python script	5 × 5 × 3 = 75
Clean and reconcile data	Excel, 3 hrs, senior analyst	Accelerate	ChatGPT Code Interpreter: "Clean this dataset — remove duplicates, standardise date formats, flag outliers beyond 2 standard deviations"	5 × 5 × 4 = 100
Calculate KPIs	Excel formulas, 1 hr	Automate	Pre-built Python script	3 × 5 × 4 = 60
Write variance commentary	Manual, 4 hrs, painful	Augment	Claude: "Analyse these variances vs budget. For each KPI, write 2–3 sentences explaining the likely driver and whether it's a concern or expected."	5 × 5 × 5 = 125
Format the report	PowerPoint, 2 hrs	Accelerate	Gamma / Beautiful.ai + template	3 × 5 × 4 = 60
Review and approve	Email chain, 1–3 days	Accelerate	AI-generated summary email with key findings for faster sign-off	4 × 5 × 3 = 60

Result: This team reduced their monthly reporting cycle from 14 hours of manual work to 4 hours of AI-augmented work — a 71% time saving. The highest-priority item (variance commentary) was tackled first because it combined the most time saved with the easiest implementation.

Workshop Facilitation Tips

Start with one process. Don't try to canvas the entire department. Pick the workflow that causes the most pain or consumes the most time.
Include the people who do the work. Managers often underestimate step complexity. The person who actually does the reconciliation knows where the real bottlenecks are.
Time-box each column. Give 10 minutes per column. Perfection is the enemy — you can refine later.
Use sticky notes or a digital whiteboard. Physical movement keeps energy up. Miro or FigJam work well for remote teams.
Score priorities as a group. Individual scoring then group discussion surfaces hidden assumptions about what matters most.
Pick the top 3 and pilot. Don't try to implement everything. Choose three high-priority items and run a 2-week pilot.

Priority Scoring Guide