Training an AI Interview Agent might sound complex, but ForceHQ simplifies the process into a few intuitive steps. By the end of this tutorial, you will have a functional, highly intelligent agent ready to screen candidates for your open roles.
Step 1: Define the Agent's Purpose
Before touching any configuration, be clear about what this agent needs to accomplish. Is it a brief 15-minute initial screen for cultural fit? Or a deep 60-minute technical dive into system architecture? The purpose dictates everything else.
Step 2: Create the Agent Profile
Go to the AI Interviewer dashboard and click Create Agent.
- Name: Give the agent a clear name, e.g., "Senior Frontend Engineer Screener".
- Persona: Select a persona that matches the role. For engineering roles, the "Technical Lead" persona is usually best as it naturally asks follow-up questions about implementation details.
- Voice: Select an AI voice model. We offer various accents and tones. Preview them to find one that feels professional and welcoming.
Step 3: Upload the Context (Job Description)
The AI needs to know what the job entails. In the Context section, paste the full job description. The platform's NLP engine will immediately parse this text and suggest core competencies that the agent should evaluate.
Pro Tip: If you have an internal document detailing the specific tech stack or architectural standards the candidate needs to know, upload that as supplementary context!
Step 4: Configure the Evaluation Rubric
This is the most critical step. The rubric tells the AI how to score the candidate.
- Review the auto-generated competencies from the previous step. Keep the relevant ones and delete the noise.
- For each competency, define strict scoring criteria. For example, under "React Performance Optimization":
- Score 1-2: Only mentions basic concepts, struggles with memoization.
- Score 3-4: Understands useMemo/useCallback, can identify re-render causes.
- Score 5: Deep understanding of concurrent mode, profiling tools, and complex state management optimization.
The clearer your rubric, the less hallucination or subjective scoring you will encounter.
Step 5: Provide Seed Questions
While the AI dynamically generates questions based on the conversation flow, providing 3-5 "Seed Questions" ensures certain topics are always covered. Enter these in the Question Bank tab.
Step 6: Conduct a Test Run
Never deploy an agent without testing it first! ForceHQ includes a built-in simulator.
Click the Simulate Interview button. You will enter a real-time voice call with the AI agent you just built. Act like a candidate. Give a few good answers and a few deliberately terrible answers. After the simulation, review the generated report to see if the AI accurately caught your mistakes and scored you according to the rubric.
Step 7: Go Live
If the simulation results look good, your agent is ready. Click Publish Agent. You can now start sending interview invitation links to candidates in your pipeline.