◕Candidates

For people you can put forward for positions — with profile, career history and documents.

From the candidate file, two paths lead onwards: to Signature (public profile on the website) and to profile optimisation (tailoring the CV to a specific vacancy). Both can be found in the file itself.

Via “+ Candidate” by hand — or via a LinkedIn address: the AI then fills in the rest from the public profile.

Check the result before you accept it. What is read from a profile is only as up to date as the profile itself, and the assignment of roles and skills is an assessment, not a fact.

Enlarge

The search field covers name, email, role and location — and always the entire pool, even if the view only shows the first 500.

In addition, there are filters by availability (available, partly available, not available), priority, roles and skills.

You paste in a requirements profile — at least 20 characters — and have the best matches found from your own pool. That is faster than leafing through by hand.

Two things you should know:

• Only your pool is searched. The match finds nobody you do not have.
• The result is a ranking with reasoning, not a decision. Read the reasoning; it reveals whether the model understood the requirement correctly at all.

In the candidate file, as uploaded files — CV, references, profile picture.

A note on operations that once proved expensive: uploaded files belong in permanent storage, not in the application image. If this is set up wrongly, the files appear to be there and disappear with the next rebuild. If you suspect that files are missing, report it immediately — the sooner, the more likely it is that something can be recovered from the backup.

Because the image addresses there are only valid for a limited time. An image loaded directly from the external server can no longer be reached once that period has expired — then there is an empty space, even though nothing has changed for the candidate.

That is why Aurora Hive Sales stores images locally instead of linking them. If you still see gaps, your instance is running on an older version.

A maintained list of roles and positions against which candidates are classified — available via “✦ AI roles”.

The point is consistency: without a fixed list, the same role is called something different for five candidates, and after that nobody can find it. The list can be managed in the interface; extend it instead of using free text.

Via “✉ Write to” in the candidate file. The window matches the one for contacts: mailbox, stage, template, subject, text, attachments.

For series mailings to many candidates, the same applies as for contacts: legally, they are advertising emails, with an unsubscribe link and everything that goes with it.

Candidate data is particularly sensitive: it contains career history, availability, often salary expectations, and the person concerned usually deals with you directly, so they can ask.

Three points that count in practice:

• Purpose limitation. Data collected for a specific placement may not simply be used for everything else.
• Retention period. Decide how long a profile stays without contact, and stick to it.
• Right of access. The file shows everything that is stored — that is the basis for answering an access request.

If AI is involved in the selection, the high-risk classification of the AI Act comes on top.

Via the setting on the candidate: available, partly available, not available. You can filter by it, and for published profiles it appears visibly on the website.

The catch is that it ages: an availability that nobody has touched for six months is not information but a guess. Make it a habit to update it with every contact — it is one click.

As your own assessment of how closely you work with someone — high, medium, low. It does not control anything automatically, but helps with sorting.

Keep the way it is assigned consistent in the team. A priority that everyone assigns differently is just as useful as a colour that everyone means differently.

Like a job description, not like a list of keywords. The field requires at least 20 characters; it becomes useful from a few sentences upwards.

What belongs in it: task, industry, environment (corporate group or medium-sized company), duration, location and the two or three points that are essential. What you can leave out: stock phrases from the advert that apply to every position.

Afterwards, read the reasoning for the matches — it shows whether the model understood the requirement.

Three paths lead onwards from the candidate, all from within the file:

• Profile optimisation — tailoring the CV to a specific vacancy and sending it.
• One-pager — a single-page profile that can be sent or put online.
• Signature — the public profile on the website.

Which one fits depends on the occasion: the tailored CV for a specific position, the one-pager for a conversation, the website profile for lasting visibility.

This mainly happens when a candidate is created via a profile address in addition to an earlier manual entry.

Merge the two and keep the record with the longer history — notes and documents are harder to recover than master data. Afterwards, check whether a published profile points to the right record.

For storage in the context of a specific placement, you usually rely on the pre-contractual relationship; for keeping the profile in your network permanently, you need consent.

It is mandatory in two cases:

• When the profile appears publicly — on the website or as a one-pager on the internet.
• When you pass documents on to third parties.

Record the consent in a way you can prove. A candidate pool without proof is vulnerable at the first enquiry.