The software for HR consultants

Imagine an additional colleague who knows your OktoCareer database and completes tasks you describe in plain language, while also drawing on other data sources and selected folders on your hard drive. That is exactly what an AI agent connected to OktoCareer does.

Technically, this happens via MCP, a standard interface:

• Nothing changes in OktoCareer. Your data stays where it is, and you keep working as usual.
• Your agent gets its own access with its own permissions. It only sees what you allow it to.
• You talk to your agent in the agent’s user interface. The agent talks to OktoCareer. For example: “Summarise the applications received for assignment 4711 and tell me which details are still missing from the requirement profile.”
• You decide which AI you use.

Most of all where time currently goes into routine work. A few typical examples:

On the candidate side

• Incoming CVs are read out, summarised and filed in the candidate record.
• For an assignment, a shortlist is created including the reasons why a profile fits and where it falls short.
• Rejections, follow-up emails and appointment proposals are prepared; you review them and send them off.
• Gaps in profiles are found, such as missing references or unclear career stages.
• From several candidates, a comparison overview is created for the presentation to the client.
• Creation of individual interview guides for candidate interviews.

On the client side

• Before the client meeting, a summary is created from the history, open assignments and recent activities.
• For acquisition, company lists are prepared, sorted and enriched with research results.
• From the conversation with the client, a clean requirement profile is created and, from it, a job ad text.
• New companies, contact persons and assignments are created and existing ones are supplemented.
• After a phone call, the conversation notes are filed in the right place.

Recurring tasks and appointments

• The agent works on a schedule. For example, it can check every morning whether new applications have come in for a position and report the interesting ones to you.
• Deadlines and follow-ups are kept in view, such as candidates with no response for ten days or assignments without a proposal for four weeks.
• Appointments and follow-up actions are prepared and submitted for confirmation.
• Each week, a short overview of the status of your ongoing assignments is created.

Documents and analyses

• Candidate profiles are generated as PDFs, cleanly prepared and ready to forward to the client.
• For an assignment, an evaluation is created of what has happened so far, who has been approached and how they responded.
• Metrics such as time-to-fill, hit rate or activity per consultant are compiled and output as a report.
• Lists of candidates, contact persons, companies or assignments are provided as CSV, Excel or PDF files.

The value lies in the preparation. The agent tidies up, researches, adds things up and drafts. Assessing and deciding is up to you.

You need three things: your OktoCareer access, an AI agent of your choice and the activation of the interface by us.

1. You get in touch with us and we enable the interface for your system.
2. In OktoCareer, you create an API key for the agent, defining its validity period, the permitted type of access and whose permissions should apply.
3. You store the key in your AI agent. This is a one-time step and done in a few minutes.
4. You store the consent text under Settings and Data Protection so that your candidates can decide.
5. You start with one or two tasks, such as reading out CVs, and expand after your first experiences.

You do not need your own server. You settle the costs for your AI provider directly with them; they are not part of your OktoCareer contract.

You decide yourself which AI agent you connect and where it runs. This also means you decide where the data flows. In practice, the following approaches are considered data-protection-friendly:

1. AI model in-house

You run a language model on your own server, in your own data centre or in your own private cloud. The data does not leave your infrastructure, and a processor never comes into play at all. This is the strictest option and, for sensitive candidate data, the cleanest.

2. Enterprise or business plan of an AI provider

The business-customer plans of the major providers differ significantly from the free consumer offerings. Look out for:

• a data processing agreement under Art. 28 GDPR
• the contractual assurance that your inputs are not used to train the models
• processing in the EU or at least a viable transfer instrument for third countries
• short or deactivatable retention periods for inputs and outputs

3. Connection via the provider’s API

If the agent is connected via a programming interface, the terms and conditions for developers usually apply. Here, too: conclude a data processing agreement and check whether training with your data is excluded. With some providers, this setting has to be set explicitly.

4. Only pass on the data that is needed

Via the interface, the agent accesses individual records in a targeted way, not the entire data set indiscriminately. Keep access as narrow as possible and issue separate credentials for the agent so that it remains traceable what it has done.

5. Candidate consent, built into OktoCareer

OktoCareer has its own consent for processing by AI. You formulate the text for it yourself under Settings and Data Protection; it appears to the candidate in the Privacy-Center for selection.

What matters is the effect: Without active consent, the interface does not deliver a candidate at all. Your agent therefore does not even see this record. This is not a subsequent sorting-out, but a block before the query.

6. Access is technically restricted

For the agent, you create a separate API key in OktoCareer. It can be given an expiry date and be bound to a fixed IP address, is subject to usage limits and can be restricted to reading or writing. In addition, the agent only sees what the user in whose name the key was created is allowed to see. Every call is logged.

7. The decision stays with the human

An AI agent prepares, sorts and suggests. The selection and hiring decision is always made by the recruitment consultant. Automated decisions with legal effect on the data subject are only permissible in narrow exceptional cases under Art. 22 GDPR.

Note: This overview is a guide and does not replace legal advice. You remain the controller within the meaning of the GDPR for the AI agent you deploy. Please seek legal advice for your specific use case.