Most companies make three costly mistakes with AI: they automate work that never needed it, they build agents they have to babysit, and sometimes the AI ends up costing more than paying a person to do the same job.
TIFDA is the framework I run with every client, and in my own agency, to avoid all three. You list your tasks, score each one, and it tells you what to automate first. List the sub-tasks under each task too, since most workflows are really a stack of smaller ones.
The five signals
- T · Time (1 to 5): a couple of minutes scores a 1, hours or days scores a 5. Score it relative to your other tasks.
- I · Impact (1 to 5): little effect on revenue scores a 1, a core money-making service scores a 5.
- F · Frequency (1 to 5): yearly is 1, quarterly 2, monthly 3, weekly 4, daily 5.
- D · Documented workflow (yes or no): do you have the steps written down, or examples of how it's done?
- A · Available data (yes or no): do you have real examples the AI can learn from?
T, I, and F add up to a score out of 15. D and A are a gate: if either is a No, the task drops out, because without a workflow and examples the AI has nothing to learn from and you end up babysitting it.
Score your tasks
List each task and its sub-tasks, then score them. Set T, I, and F from 1 to 5, and set D and A to Yes or No. The verdict on each row tells you what to do with it, and your ranked shortlist builds itself below.
| Task | T + I + F | D | A | Verdict |
|---|
Your shortlist
Missing a documented workflow? Don't let it stop you. Almost no company has everything written down. Have the person who does the task record their screen and narrate it, then feed that to the AI to draft the workflow. Missing data is the real blocker: without good examples, the AI can't tell great work from bad.