Input Quality

Working From Messy Inputs

Nobody hands you a clean brief. You have Slack threads, half-written notes, fragments of performance observations, and a vague stakeholder comment from a hallway. The single biggest barrier to good AI output is not technical skill, it is messy input. Output quality tracks closely with input quality.

The Four-Step Prep

Before You Paste Anything In

1

Include

The facts, observations, and context the AI needs. Be generous with relevant context, this is where quality comes from.

2

Cut

The noise that distracts or misleads: off-topic chatter, duplicate points, emotional venting that is not the task.

3

Anonymize

Strip identifying details before they go into a commercial tool: full names, employee IDs, anything personal or confidential.

4

Verify

Plan to spot-check the output, especially anything the AI states as fact. Messier input means more human review needed.

Privacy Rule (starts now)

If you would not post it on a public noticeboard, do not paste it into a public AI tool. This is the GDPR-grounded rule formalized in the final module, and it applies starting now, on every task in this course.

Common Messy Inputs

What to Do With Each One

Dana's story

Dana needs to turn a chaotic forty-message Slack thread about a slipping project into a clear status summary for leadership. Raw, the thread is unusable: side jokes, a vent about another team, two people's half-formed guesses. Dana copies it, cuts the noise and the venting, replaces names with roles, keeps the actual decisions and blockers, and then prompts for a structured summary. The output is clean because the input was. Then Dana spot-checks one date against the original thread before sending. Five minutes of prep turned a garbage-in into a usable-out.

Apply It Today

Run the four-step prep on one real input

Take one messy input from your real week (a thread, a transcript, a pile of notes) and run the four steps before you prompt: include, cut, anonymize, verify. Notice how much the cleanup alone improves the result.

Reflect

  • Which of the four steps do you skip most often, and what does it cost you?
  • What is the most sensitive thing you have pasted into an AI tool without thinking about it?