In 30 seconds

Before adding intelligence, give the workflow reliable structure: a clear digital input, named states, and a proper record of what happened. Many problems that look like AI problems are really missing structure.

Digitizing is not the same as automating. It is making the work legible to software.

Why this comes up

AI can produce convincing output on top of a messy foundation, which hides the mess instead of fixing it. When the output is wrong, the cause is often underneath: information arrives inconsistently, statuses live in people's heads, and there is no record of who did what.

Fixing the foundation first makes everything above it more reliable, and sometimes removes the need for AI at all.

What "digital enough" means

You do not have to replace every system or buy new software. You need three things to be reliable.

1. A clear input

The work should start from a defined digital event carrying the information the workflow needs. A reliable intake form often does this. It replaces "however the request happens to arrive" with a consistent starting point.

2. Visible states

Each item should have a named state that says where it is and what happens next. States turn a pile of work into a manageable flow and make handoffs real.

3. A record of what happened

The workflow should leave a trail: what arrived, what was decided, what changed, and by whom. This is what makes a process auditable and what lets you review failures later.

Common problems that look like AI problems

Many "AI problems" are really:

  • missing or inconsistent data
  • no unique identifiers, so records cannot be matched
  • unclear status fields
  • documents with no owner
  • emails used as a database
  • spreadsheets used as a workflow engine
  • verbal approvals with no record
  • duplicate sources of truth

If you see these, fixing them is usually the highest-value work you can do, with or without AI.

A small example

A team processes membership applications by email. Applications arrive in different formats, status lives in who has replied, and there is no record of decisions.

Digitizing means: a simple intake form so every application has the same fields, three clear states (received, reviewed, decided), and a record of each decision with a reason. After that, an AI step to pre-check applications becomes easy to add and easy to trust, because it has consistent input and a place to put its result.

What this does not solve

Digitizing does not decide whether AI belongs in the workflow, and it does not replace good judgment about what to change. Some manual steps are manual for good reasons. The goal is better work, not maximum automation.

Good enough to move on when

  • The workflow starts from a consistent digital input.
  • Every item has a clear, visible state.
  • Decisions and changes leave a record.
  • The obvious data problems have been addressed.