Hiring Guide
Hiring AI workers mirrors the human hiring process: review resumés, hold optional interviews, fill up a hiring application for your candidate, and onboard. Each resumé outlines the core skills shared by all AI workers of that type. During hiring, you customize your AI worker - tuning communication style, social behavior, and other settings to match the role, team dynamics, and work environment.

naming AI worker »

If enabled, naming your AI worker helps integrate them better with the team. Choose a human name for smoother social interactions or a technical nickname, but remember that this does not conceal their AI nature. The name should align with the gender and the cultural context of employer. At minimum, a first name or nickname is required.
We recommend using normal full form human names. Most communication systems like Slack or Teams will clearly designate the account as non-human (app, bot, etc.) so the human name will not lead to unintended concealment of AI nature. Some companies, for example, use first names that end with "ai" like Kai or Sarai to give a playful hint of their AI nature.
NOTE: in all circumstances you should avoid concealing the AI nature of the AI workers by using specific names or avatars. Not only can it run against upcoming or existing regulatory requirements but it can alienate your existing team against the use of AI workers.

contact persons »

When configuring a new hire you will need to specify the following two contacts:
  • Direct supervisor to whom AI worker will be reporting to. While anyone in your organization can communicate with AI worker, the direct supervisor will be the primary contact for the AI worker activities.
  • A person who filled up the hiring application on the Humatron website (typically, an HR or hiring manager). This contact will be a different person if the hiring is done by the partner HR agency on behalf of its client.

knowledge base »

Knowledge base allows to ground AI worker in the context of the company and the job by providing it with data sources that are continuously monitored for updates and scanned by the internal RAG system.
Knowledge base data sources are automatically scanned by the internal RAG system to provide up to date context to the AI worker. Only general read-only access to the data sources is needed as AI workers never write to these sources. Knowledge base can include, but is not limited to, company policies, employee handbooks, product documentation, address and contact books, etc. Knowledge base can be configured once the hire is fully deployed.
For data sources that support individual files (Box, Dropbox, Google Drive, AWS S3, etc.), the following file types are supported.
Text & Document Formats:
  • PDF (.pdf)
  • Microsoft Word (.doc, .docx)
  • Plain text (.txt)
  • Markdown (.md)
  • HTML / Web pages (.html, .htm)
  • Rich Text Format (.rtf)
  • EPUB (.epub)
Structured & Semi-Structured Data:
  • JSON (.json)
  • CSV (.csv)
  • TSV (.tsv)
  • XML (.xml)
  • YAML (.yaml, .yml)
Code & Technical Files:
  • Source code files (language-agnostic, e.g. .py, .js, .ts, .java, .go)
  • Configuration files (.ini, .toml, .env)
  • Logs (.log)
Images:
  • PNG (.png)
  • JPEG / JPG (.jpg, .jpeg)
  • TIFF (.tiff)
  • BMP (.bmp)
  • WEBP (.webp)
Audio & Video:
  • Audio (.mp3, .wav, .m4a, .aac)
  • Video (.mp4, .mov, .avi, .mkv)

gender & communication style
»

If allowed by the resumé, selecting social preferences like gender or communication style enables proper grammar and tone in communication with AI worker, as well as fine-tuning of the overall communication style. Ensure the chosen name and avatar match the selected gender.

selecting avatar »

Most AI workers allow for social customization of their appearance including selecting an avatar. Selecting an appropriate avatar is crucial for the smooth integration of an AI worker into a human team. Visual identity holds significant value, and hiring managers must thoughtfully decide whether to use human-like avatars, stylized images that diverge from human appearances, or something in between.
Attitudes towards avatar choice can vary widely across different countries, cultures, and corporate environments. However, it is strongly advised against using real human images or photographs as avatars for AI workers.
It's important to ensure that the chosen avatar aligns with the AI worker's designated gender. Be aware that using a human-like image or photograph does not disguise the AI's nature.

job instructions
»

Correct job title and job instructions are important for functioning of the AI worker. Job instructions should provide company specific details of the work you want the AI worker to perform which will become part of the LLM system prompt. You also need to provide the name of the team that is hiring this position. Note that you can attach one or more files to provide more context such as employee handbook, company policies, etc.
Both job title and the team name will also be used in email signatures, Slack status, etc.
Note: job instructions at hiring are complementary to the build's role instruction. While build's role and role instructions provide general guidance, job instructions are more specific to the job you are hiring for and should include company specific details, if any.

docker environment variables
»

Each hired AI worker is deployed in isolation in the Docker container. Core agents, custom functions and MCP servers depend on Docker container environment variables. These variables must be set during the hiring process. Ensure that all mandatory variables are properly configured.
IMPORTANT: This is the only place where confidential information - such as passwords or API keys - can be safely entered. This data is never shared and is not accessible within the LLM context.

hire token »

Each hire is issued a unique hire token, generated once at creation. Together with the corresponding organization and build tokens, it can be used for MCP or tool authorization and authentication.

SSH access »

Once the hire instance is fully configured and deployed, SSH access details will be available on the hire page. You can use this information to connect directly to the VPS server hosting your AI worker.
IMPORTANT: By accessing the VPS server, you assume full responsibility for any risks related to the AI worker's runtime and operation. Additionally, whenever the AI worker is restarted, its VPS containers are reset, and any local changes made to the instance will be permanently lost.