batteriesincluded.com · Questions & Answers

Our files and spreadsheets are messy. Do we need to clean up all our data before starting our first AI project?

No. Clean up the inputs required for one useful workflow before attempting a company-wide data cleanup. Messy information is a reason to narrow the first project, not automatically a reason to postpone all AI work.

Choose a recurring task and trace the information it actually uses. Identify the source of each required field, who maintains it, and what happens when it is missing or contradictory. If employees cannot agree on the correct source, resolve that disagreement before asking AI to interpret it.

For the initial scope:
- Use a bounded collection of current, relevant records.
- Remove duplicate or obsolete inputs that would change the result.
- Identify which information the tool must not access.
- Define how incomplete records reach a person for review.
- Validate a small sample of outputs against the original records.

The useful test is whether the workflow can produce checkable results from its actual inputs. A polished demonstration using unusually clean examples does not establish readiness for everyday use.

If source information changes frequently, assign responsibility for keeping it current. Otherwise, an initially useful application can become unreliable even if its code never changes.

A Batteries Included project should begin with that bounded business request. Lovable or Replit can support the custom build, but neither removes the need for reliable inputs. Expand the dataset only after the small workflow produces useful results with an acceptable review burden.

Category: First AI Project

← All questions