Creating and Deploying a Sage

How to design, build, test, and deploy a purpose-built AI screening agent configured for your role, culture, and evaluation standards.

Overview

A Sage is a purpose-built AI screening agent configured to evaluate candidates for a specific role within your organization. Each Sage inherits its foundational capabilities from a Sage Mentor, a base model trained in structured behavioral interviewing, and is then shaped by your inputs to reflect the role, culture, and evaluation standards you define.

The lifecycle of a Sage follows a deliberate sequence: design, build, test, deploy. Design determines what the Sage evaluates. Build translates your inputs into a functioning interview agent. Testing lets you experience the screening firsthand. Deployment makes it operational. Once deployed, a Sage is locked; every candidate is evaluated against the same criteria.

Designing the Evaluation Framework

Role Context

Role Context tells the Sage what it is evaluating for. You provide:

• Sage Name: How the Sage introduces itself to candidates. Choose something that fits the personality you want projected.
• Voice and Language: The text-to-speech voice and language the Sage uses when conducting screens. This is a functional selection, not cosmetic.
• Role: The position title. This anchors the Sage’s understanding of the job domain.
• Role Level: The seniority or experience tier. The Sage calibrates interview depth accordingly.
• Job Description: The most consequential input in the entire process. Describe the role in detail: responsibilities, expectations, team dynamics, technical requirements, success criteria.

Tip

Write the job description as if you are briefing a senior colleague who will conduct the interview on your behalf. That is exactly what you are doing. A vague, copy-pasted description produces a generic interview. A precise, honest description produces questions that surface real signal.

Cultural Alignment

Cultural Alignment shapes how the Sage interviews and evaluates, beyond technical fit:

• Intro Question: The opening question the Sage asks every candidate. You also provide the reasoning behind it, so the Sage understands why this question matters.
• Evaluation Guideline: Controls rigor. Options range from Flexible through Relaxed and Balanced to Strict. Each level automatically calibrates how high the bar is for a Moderate or Strong result. The guideline is locked once the Sage is built.
• Tone: Select up to five personality traits (e.g., Professional, Friendly, Direct, Encouraging, Assertive, Warm, Casual, Formal). These shape the Sage’s conversational style during the interview.
• Additional Instructions: Free-text field for anything the standard fields do not cover. Use this for role-specific screening rules, topics to emphasize, or behavioral signals to prioritize.

Tip

Strong cultural inputs are specific and intentional. “Be friendly” is a tone selection. “Prioritize candidates who demonstrate ownership over projects, not just participation” is an instruction that changes what the Sage listens for.

Review and Build

Review

Before build begins, you review a complete summary of every input you have configured: role context, cultural alignment, tone, evaluation guideline, and instructions. This is your last opportunity to adjust. Once the build starts, these inputs are committed.

Build

Once confirmed, the system enters an automated build phase that typically takes approximately ten minutes. During this process, the Sage Mentor uses your inputs to construct:

• A tailored interview structure with contextually relevant questions
• An evaluation framework with criteria aligned to your role and culture inputs
• Adaptive follow-up logic that responds to candidate answers in real time
• An evaluation rubric calibrated to your evaluation guideline

Note

The build runs in the background. You can close the page and continue working; you will receive an email notification when the build completes. “Build complete” means the Sage is ready to conduct interviews. It does not mean the Sage is deployed or visible to candidates. Nothing happens until you explicitly move forward.

Validation and Testing

Testing is optional but strongly recommended. It is the only way to experience the screening as your candidates will.

When you initiate a test, the system sends you an invitation email identical to what candidates receive. You click the link, enter the screening, and experience the full interview: same questions, same evaluation, same interface. There is no abbreviated or simulated test mode; each test link is unique and single-use, just like a candidate’s.

Test screens are functionally identical to live screens with one distinction: they are flagged internally and excluded from all reporting, statistics, and candidate dashboards. Your test results exist for your review only.

If you choose to test, evaluate the following:

• Signal Quality: Are the questions surfacing information that actually helps you make a decision about a candidate?
• Alignment: Does the Sage’s focus match the priorities you defined in your inputs?
• Tone: Does the conversational style feel appropriate for the role and your organization?
• Competency Coverage: Is the Sage probing the areas that matter, or spending time on irrelevant topics?

Caution

If testing reveals problems, resolve them before deploying. Once a Sage is deployed, its evaluation framework is locked and cannot be changed.

Deployment

Activation

Deployment is a status change. During deployment, you choose one of three states:

State	What It Means
Active	The Sage is live. You can immediately begin creating screens and inviting candidates.
Testing	Available for internal validation only. No candidates are invited. Team members can take the screening and provide feedback before you go live.
Draft	The Sage is saved but not operational. Return to it later.

At deployment, you configure or confirm the email invitation template that candidates receive. This template supports dynamic variables for personalization and can be customized per Sage or default to your organization-level settings.

Immutability After Deployment

Deployed Sages cannot be modified. This is intentional. Allowing changes to a live Sage would mean candidates screened at different times could be evaluated against different standards. That compromises the integrity of the process and makes it impossible to fairly compare candidates. By locking the Sage after deployment, every candidate is evaluated against the exact same baseline.

The only changes you can make to a deployed Sage are its voice and language; the evaluation framework, evaluation guideline, questions, rubric, and instructions remain fixed. To change any of those, build a new Sage.

Managing a Deployed Sage

From the Sage detail page, you can:

• Toggle status: Switch between Active and Inactive. Setting a Sage to Inactive immediately cancels all draft and invited screens, refunds the associated credits, and prevents new screens from being created. Screens already accepted or in progress are allowed to finish.
• Change voice or language: Update the text-to-speech voice or the screening language without affecting the evaluation framework.
• Create screens: Invite candidates directly from the Sage detail page.
• Review performance: The detail page surfaces screening statistics, role alignment data, and a full grid of associated screens.

When to Create a New Sage

Because deployed Sages are immutable, iteration means creating a new one. Create a new Sage when:

• The role has changed in scope or responsibilities.
• Testing or screening results reveal that your inputs need adjustment: different evaluation guideline, different tone, different instructions.
• You need to screen for a different position.
• You want to experiment with a different approach while preserving the current Sage’s data.

Note

Setting the original Sage to Inactive retires it cleanly. Its historical data, completed screens, reports, and results remain accessible. The new Sage starts fresh with its own screening history.