AI Disclosure

Crestento is an AI writing assistant. It helps managers draft performance review prose from bullet points and reference materials they provide. It is not an automated decision system.

What the AI does

• Generates suggested review language based on the structured input (context, accomplishments, strengths, growth areas, goals, overall, sensitive notes) the manager provides.
• Calibrates tone and language to the role chosen from our role library.
• When attached, considers reference documents (job descriptions, performance plans) as context.
• When configured, mirrors the cadence and vocabulary of the manager's own writing samples saved in their profile.

What the AI does not do

• Evaluate, score, or rank employees. Crestento does not produce performance ratings or comparative scores about any individual.
• Recommend hiring, firing, promotion, compensation, or disciplinary action. The AI is explicitly instructed to refuse such recommendations in the system prompt. If they appear in your bullets, the AI will render them as draft language but make no judgments of its own.
• Invent content. The AI is instructed to use only information you provide in the form fields and attachments. If a section is left blank or marked "skip", the AI will acknowledge the gap rather than fabricate.
• Persist your prompts for training. Our AI provider (Anthropic) does not retain inputs sent through their API for model training.

Human-in-the-loop responsibility

The user — you, the manager — is the author of any review you save, send, or act upon. Crestento drafts language; you read, edit, and approve it. You are solely responsible for the content of any review used in employment decisions.

Models in use

Crestento drafts reviews using leading Anthropic Claude models via the Claude API. We may shift between models as Anthropic releases updates, or use a higher-capability model on premium tiers; we'll note any material change here.

Bias and accuracy

Large language models can produce biased, inaccurate, or inappropriate output. Crestento's prompt structure (role calibration, structured input, anti-fabrication rules) is designed to mitigate these risks, but it does not eliminate them. We encourage every user to:

• Read drafts carefully before saving or sending.
• Edit any phrasing that doesn't reflect your honest assessment of the employee.
• Avoid relying on the draft as evidence of performance — the only evidence is what you put in the bullets.

Regulatory positioning

Crestento is a writing tool. It is not a "high-risk" AI system under the EU AI Act (no employment decisions are made or scored by the system). It is not subject to NYC Local Law 144 (no automated employment decision tooling). We treat employee data with the same care as any sensitive workplace data; see our Privacy Policy and Data Processing Agreement.

Questions

If you're evaluating Crestento for compliance review or have questions about how the AI behaves in a specific context, get in touch.