Design an AI workflow
AI Workflow Design Framework
The AI Workflow Design Framework structures an automated marketing process in eight steps: trigger, input, AI task, validation, human review, action, output and measurement, so that each step has an owner and a check before anything runs on its own.
The model
- 1Trigger
- 2Input
- 3AI task
- 4Validation
- 5Human review
- 6Action
- 7Output
- 8Measurement
Developed by Markaigen · Updated
What each part means
- Trigger
- The event that starts the workflow: a new lead, a keyword opportunity, a date, a form, a threshold.
- Input
- The data and context the AI step receives, and where it comes from.
- AI task
- What the model does with the input: classify, draft, summarise, score, extract or decide.
- Validation
- Automatic checks on the AI output: format, required fields, banned words, limits and confidence.
- Human review
- Who looks at what, and when: always, by sampling, or only on exceptions.
- Action
- What the workflow does with an approved result: publish, send, update, create a task.
- Output
- The result as the next person or system receives it, in the format they need.
- Measurement
- How you know it works: volume, error rate, time saved, outcome metric and review load.
How to use it
- Draw the eight steps before you open an automation tool; the tool should follow the design.
- Put human review before any action that cannot be undone.
- Start with always-review, then move to sampling once the error rate is known.
Draft a workflow outline
Fill in the eight steps. The outline is ready to paste into a workflow tool or a brief for the person who builds it.
Answers are saved in this browser so you can return to them.
Download or e-mail this framework
Get a branded PDF or Word document, with your answers or as an empty template to fill in.