In 30 seconds

A workflow can only use information it can actually reach. Before designing, check that each step's required data exists, is reliable, and can be accessed. Missing or unreachable data is one of the most common reasons AI projects stall.

Why this comes up

It is easy to design a clever workflow on paper and discover, weeks later, that the data it depends on is trapped in a system nobody can connect to, or does not exist in usable form at all. Checking availability early turns that late, expensive surprise into an early, cheap finding.

For each step, ask three questions

Does the information exist?

Sometimes the data a step needs is simply not recorded anywhere. "Route by customer priority" needs a priority field that may not exist. If it does not, someone has to create and maintain it first.

Is it reliable?

Data that exists can still be wrong, incomplete or stale. If half the records are missing the field, or the values are inconsistent, the step will fail for those cases. Check quality, not just presence. Confirm there is a source of truth.

Can the workflow reach it?

Even reliable data is useless if the workflow cannot access it. Check whether the system holding it has a usable integration or API. Some systems are closed, and that is a hard constraint, not a detail.

Distinguish read from write

Be clear whether each step needs to read data or also change it. Reading is usually easier and safer to arrange than writing. A workflow that only reads may be quite feasible even where writing back would not be. See read access vs write access.

A small example

A team planned an assistant to answer order-status questions. The status lived in a warehouse system with no way for other software to read it. That single availability gap decided the project: either the warehouse system had to expose the data, or the assistant could not work. Finding this in week one, not week ten, saved months.

What this does not solve

Confirming data is available does not make it correct or keep it current. Availability is the entry condition; quality and freshness are ongoing work. And available data still has to be handled with the right permissions and validation.

Good enough to move on when

  • Each step's required information is listed.
  • You know where each piece lives and whether it is reliable.
  • You have confirmed the workflow can reach it.
  • Read and write needs are distinguished for each source.