rAlph – Bulk changes in large organisations

rAlph is only available when AI is enabled for the tenant.

With twenty roles it makes no difference how rAlph fetches its data. With three hundred it does. This page explains what rAlph does when you ask it to do something for everything – and how you can tell that it really saw all of it.

Two ways of getting data

rAlph has two tools, and they answer different questions:

Search – “Find the Backend Engineer role.” Search returns a page of matches, not necessarily all of them. That is intentional: for a targeted question it is enough, and the answer stays short.
List completely – “Assign all roles to their circles.” For this rAlph fetches the complete list of an entity type: the name, type and identifier of every entity, without descriptions or attributes. That is exactly why several hundred entries still fit into a single working step.

You do not have to steer this. rAlph decides for itself – if your request says “all”, “every” or “each”, it fetches the complete list.

How to see what it did

Above every answer there is an expandable line listing the working steps (e.g. 10 actions). Expand it and you see which tools ran:

“Complete list fetched” – it loaded the entire set.
“Entities searched” – it fetched one page. If more is waiting, it fetches the next page on its own.

When rAlph states in its text how many roles exist, that is not an estimate: it is given the real count.

When rAlph gives no overall count

Sometimes rAlph will say something like: “I loaded 120 roles; I cannot tell you how many there are in total.” That is not a failure – it is deliberate.

If entities are not visible to you (see Visibility), an overall count would reveal how many of them exist. So in that case rAlph is given no number at all – and it does not guess one. What it names is then the number of entries it actually holds.

Rule of thumb: A number from rAlph is correct. A missing number is about visibility – not about something having gone wrong.

The upper limit

The complete list is capped at 2,000 entries. When the cap is reached, rAlph says so explicitly and suggests narrowing the scope – for instance to one entity type.

The reason is not the database but the working memory of the conversation: every row rAlph reads occupies context and costs money. An honest limit with a hint is better than an answer that breaks off halfway through a bulk operation.

If you need more, work in stages: first the roles of one division, then the next.

What rAlph does not see

Even with a complete list, three things stay out – each on purpose:

Modules that are not released. An administrator decides under AI settings which modules the AI may access. Anything locked is missing from the list – but rAlph says so explicitly instead of claiming “not found”.
Entities without a visibility grant. rAlph only sees what you may see.
Templates (unassociated entities, see Templates (unassociated entities)). They are building blocks, not organisational reality.