Will it work with the systems we already have?

Asked precisely: will the AI integrate in existing software, or will you create a separate application?

It is the question that decides adoption, and people underrate it. Another client asked the sharper version: have you integrated this into a system before with other customers? Always ask that. Ask for the specific system.

The default should be: into what you have

Every new application you introduce is another login, another place to look, another thing to train people on. If the work happens in a system your team already lives in, the tool should meet them there.

I have built things that sync back into the spreadsheet platform the client was already using, precisely because asking them to abandon it would have killed the project. Meeting people where they work is not a compromise, it is the design.

What determines whether it is possible

• Does it have an API? Most modern systems do. This is the easy case.
• Does it export? A scheduled export is less elegant but works, and is often enough.
• Is it on-premise and old? Usually still possible, via the database directly, but the conversation gets longer.
• Is it locked by the vendor? Some suppliers deliberately make data hard to get out. This is worth knowing early, and it is worth remembering at renewal.

The honest caveat

Integration is where estimates go wrong, in both directions. Sometimes a system I feared turns out to have a clean API and takes two days. Sometimes something that should be simple has an undocumented quirk that eats a week.

This is why I want access in week one. Not to start building — to find out which one we are dealing with while it is still cheap to know.