“Let's do it with AI” is not yet a project requirement. It may conceal a real problem, a time-consuming process or simply enthusiasm for a new tool. A short intake form makes the team define the desired change before choosing a model, supplier or prompt.
Start with the decision, not the technology
First ask which decision or action should improve. Record who currently does the work, who uses its output and the consequences of error. Without clear ownership and measurable friction, the idea is not ready for automation.
Map inputs and desired outputs
Describe incoming data, formats, frequency and exceptions. Show two good and two poor output examples. “Write better” cannot be tested; “identify three next steps and return them in defined fields” can.
Set limits on acceptable error
Failures have different consequences. A weak draft can be corrected, while an incorrect financial instruction may cause real damage. The form should capture likelihood, severity, detectability and the person approving any irreversible action.
Check permissions and sensitivity
Identify personal information, trade secrets, customer material and restricted content. Record storage locations, access and whether data may be sent to the selected provider. If these answers are missing, pause the pilot.
Define the smallest useful experiment
Select a limited sample, baseline and deadline. Measure quality, time, cost and the proportion of cases needing human intervention. The pilot's outcome is not “we liked it” but a decision: proceed, modify, narrow or abandon.
Leave room to say no
A sound intake system rejects ideas lacking data, ownership, safe review or sufficient repeatability. This does not obstruct innovation. It protects the team from impressive solutions that add more work than they remove.
The AI intake form is DigitalNow's original evaluation framework.