Most automation projects fail for boring reasons: the task was never clearly defined, nobody owned the output, or the data was not where people thought it was. These five questions catch most of those problems before any work starts.
1. Does the task repeat?
A job you do once a year is rarely worth automating. A job that happens daily or weekly, and takes more than an hour each time, usually is. Count the hours honestly across your team.
2. Can you describe a good result in writing?
If you can write down what "done well" looks like, with two or three real examples, an agent can aim at it. If the standard lives only in one person's head, write it down first. That exercise is valuable even if you never build anything.
3. What happens when it is wrong?
Every system makes errors. Ask what the worst realistic error would cost and whether someone would notice it before it did damage. Low cost and easy to spot is the ideal starting point. High cost and invisible means a human must stay in the loop.
4. Where does the data live, and who may see it?
List the tools involved (email, spreadsheets, a CRM, a shared drive) and decide which ones the agent may read and which it may write to. Give access to the minimum needed and keep it revocable. If personal data is involved, check the rules that apply in your country before you begin.
5. Who approves the output?
Name one person. Agents should prepare, and people should approve, at least until the agent has earned your trust on that specific task. Over time you can let low-risk tasks run without review.
If you can answer all five questions in a few sentences each, you are ready for a pilot. If not, you have just found out what to fix first.
We use these same questions on our discovery call. Book one and bring a task you are curious about.
Have a task like this in mind?
Describe one task that takes too much of your team's time. On a 30 minute call we will tell you honestly whether an agent can do it, and what it would take.