Creating AI agents and their competencies

In Swarmica, you can create AI agents that perform routine tasks using MCP tools. Special knowledge in the field of AI is not required: the administrator sets the agent's name and selects its competence, after which the agent is ready to work.

How to create an AI agent

1. Open the employee management section and select the creation of a new AI agent.
2. Specify the name under which the agent will be displayed in the system.
3. Select the competence.
4. Save the agent and start the desired task or assign it to the appropriate process.

The agent's status and progress of tasks are displayed in the interface. ChatGPT and YandexGPT providers are available for AI agents to work. ChatGPT is usually more advantageous in terms of token consumption, while YandexGPT is suitable if data storage in Russia is more important.

Competencies of AI agents

QA evaluator

Checks applications and evaluates the work of the support employee according to service quality criteria. The agent assigns scores and leaves detailed comments on why the criterion was met or not met.

AQI evaluator

Analyzes knowledge base articles, identifies deviations from corporate standards, style, and information security requirements.

LAI evaluator

Checks compliance with the KCS process: whether the employee searched for information in the knowledge base when handling the request and documented new knowledge correctly.

Processing appeals against assessments

The agent analyzes the employee's objections to QA, AQI, or LAI scores. If the arguments are justified, it corrects the score; otherwise, it explains the decision and forms a final comment.

Article editor

Brings verified articles to the company's content standard and style: corrects grammar, standardizes formatting, and publishes the finished material.

SEO/GEO/AEO article optimization

Fills in meta tags and JSON-LD markup of published articles to make the knowledge base easier to find by search engines and generative models.

Article translator

Translates published articles that have only one locale, preserving the formatting.

How to use competencies

The AI agent performs the first pass, and the employee remains the arbitrator: checks disputed cases, makes decisions, and uses the results to train the team. This approach allows you to increase the quality control sample, receive feedback faster, and free up time for experienced employees to handle complex cases and mentoring.

In conjunction with quality control rules, the process looks like this: the rule creates assessments, the AI agent conducts the check, and the manager analyzes the result and makes a decision.

The ability to create AI agents depends on the enabled functionality __ai_mcp_agents.

Sources:

• Swarmica 6.0: AI agents, ITIL, RCA, and interface customization
• Update v6.1.x