Inputs

What goes into the solution (data, material, etc.)

Inputs are everything the solution needs to receive before it can do anything useful: the data, files, text, context, or decisions that must arrive first.

AI is not intuitive about what it is missing. If an input is absent, the tool will often proceed anyway, producing output that looks confident but is built on nothing. Knowing what goes in is not just a design detail, it is a prerequisite for getting a result you can trust.

Inputs for AI solutions often mix the obvious and the invisible. A customer email is an obvious input. The context about what that customer has bought before, or the tone policy your team follows on replies, is invisible unless you make it explicit. Most of the work in defining inputs is surfacing the invisible ones.

It is also worth asking where each input comes from and whether it can actually reach the solution. An input that lives in someone's head, or in a system with no integration, is a blocker that needs solving before building begins.

Make it visibleList every input the solution needs. For each one, write where it currently lives (a file, a database, someone's head, a chat thread) and whether a tool can reach it today. Treat any input without a clear source as a blocker to resolve before building.

Why AI needs this

Each part of your work matters to AI in a specific way. Some of it is context a tool needs before it can help, some of it is work a tool can take on, and some of it is judgment that should stay with you.

Preventing hallucination

When required context is missing, AI fills the gap with something plausible rather than stopping. Knowing the inputs lets you make sure nothing critical is absent when the solution runs.

Reachability check

An input that exists but cannot be accessed by the solution is as useless as no input at all. Listing inputs forces an honest check of where each one lives and whether AI can reach it.

Reproducibility

The same inputs should produce similar, trustworthy outputs. If you cannot name the inputs precisely, you cannot explain why the solution produced one result on one day and a different one the next.

Questions to explore

Use these on your own or in a group. There are no right answers, only better conversations.

What does the solution need to receive before it can do anything, and where does each of those things come from?
Which inputs are already in a system or file a tool can reach, and which exist only in a person's head or a locked tool?
What is the minimum set of inputs for the solution to produce something usable?
Are any of the inputs inconsistent, messy, or partial in ways that will affect what comes out?
Who is responsible for making sure each input arrives in the right form at the right time?

Readiness traps

Underspecified inputs are one of the main reasons AI outputs feel unreliable: the tool did its best with what it had, which was not quite enough.
Context inputs (policies, rules, history, tone) are easy to forget because they are not files or data: they live in people's heads and need to be captured and handed over explicitly.
Assuming an input will be clean and complete is almost always wrong. Build in a check for missing or malformed inputs early, or the solution will fail in unpredictable ways.