Human-First Policy

Effective Date: January 8, 2026

Applies To: All clients, operators, contractors, and partners of Zeitra.AI

Purpose: We automate repetitive work — not human creativity, empathy, or final moral, legal, or safety decisions. Where a task must stay human, we build systems that make the human faster, better informed, and more capable — never systems that remove them from the equation.

Quick Definitions

  • HOTs (Human-Owned Tasks): Tasks that must stay under human control. These include anything requiring final creative judgment, empathy-driven interaction, moral or legal reasoning, hiring/firing authority, high-stakes safety decisions, or any task the Client designates as human-owned. HOTs can still be assisted by automation — the restriction is on full replacement, not on support.
  • ATs (Automatable Tasks): Repetitive, deterministic, or rules-based steps that can be reliably executed by a system. These include data routing, formatting, monitoring, scheduling, drafts for review, and similar operational work.
  • HOL (Human Ownership List): A short checklist created during discovery that defines exactly which tasks stay human for a given engagement. The HOL is approved by the Client before any build work begins.

The Rules

  1. We do not fully automate HOTs. No exceptions without a formal process.
  2. We actively build systems that assist, support, and expand human capability for HOTs — as long as a human remains the final decision-maker.
  3. We will automate ATs to save time, reduce errors, and free humans for higher-value work.
  4. All HOTs must be listed in the HOL and approved by the Client before work starts.
  5. If a Client wants a HOT fully automated, they must submit a Change Request — Zeitra.AI will assess risk and require human-approval gates, immutable logs, and a rollback plan before any work proceeds.

The Line: Replacement vs. Assistance

This is the most important distinction in this entire Policy.

We will never build a system that removes a human from a decision they should own. But we will absolutely build systems that make that human's job easier, faster, and more informed.

The line is simple: does a human still make the final call?

  • A system that drafts a customer complaint response for a human to review, edit, and send — that's assistance. We build that.
  • A system that sends the complaint response on its own with no human review — that's replacement. We don't build that.
  • A system that surfaces candidate resumes, scores them against criteria, and presents a shortlist for a hiring manager — that's assistance. We build that.
  • A system that rejects applicants automatically with no human review — that's replacement. We don't build that.
  • A system that generates a draft estimate based on job details for an owner to review and adjust — that's assistance. We build that.
  • A system that sends a binding quote to a customer without the owner seeing it — that's replacement. We don't build that.

If the human is still in the loop, still reviewing, still approving — we're in. The moment the system acts alone on something that requires judgment, empathy, or accountability — we're out.

What We Will NOT Fully Automate

The following categories represent tasks that require human judgment, empathy, creativity, moral reasoning, or final decision-making authority. These are always classified as HOTs unless an approved exception is in place.

We will build systems that assist with every category below — drafting, organizing, surfacing information, preparing options, flagging issues, and accelerating the human's workflow. The restriction is on removing the human from the final decision, not on making them better at it.

Creative & Brand

  • Final ad creatives, logos, brand identity assets, or finished design work
  • Final written copy that represents the Client's voice without human review
  • Finished music, video, film, or multimedia content
  • Final brand messaging, positioning, or campaign direction decisions

Empathy & Human Connection

  • Counseling, therapy, or trauma response interactions
  • Customer interactions that require emotional sensitivity, de-escalation, or personal judgment
  • Bereavement, crisis, or hardship-related communications
  • Relationship-driven sales conversations where trust is the primary factor
  • Personalized customer recovery, retention, or complaint resolution conversations
  • Any interaction where the recipient would reasonably expect to be communicating with a human

Legal & Compliance

  • Final legal advice, legal opinions, or compliance determinations
  • Sentencing, disciplinary, or enforcement decisions
  • Contract execution or binding legal commitments on behalf of the Client
  • Regulatory filings or submissions that require professional sign-off

Employment & People

  • Hiring, firing, or any final employment decisions
  • Performance reviews, disciplinary actions, or promotion decisions
  • Compensation, benefits, or pay-related determinations
  • Workplace safety judgments that affect employee well-being

Medical & Safety

  • Clinical diagnoses, prescriptions, or treatment plans
  • Triage decisions that affect patient safety
  • Mental health assessments or interventions
  • Emergency response decisions where human life or safety is at stake

Financial & High-Stakes

  • Final pricing decisions that require contextual or relational judgment
  • Fraud adjudication or final dispute resolution
  • Insurance claim approvals or denials
  • Credit, lending, or financial eligibility decisions

General

  • Any task the Client marks as a HOT in the HOL
  • Any decision where the wrong automated output could cause irreversible harm, reputational damage, or legal exposure
  • Any task where a failure would go undetected without human oversight

What We WILL Automate

The following categories represent repetitive, deterministic, or rules-based tasks that are reliably executable by a system. These are always classified as ATs unless the Client designates them otherwise in the HOL.

Data Management

  • Data capture, entry, cleaning, formatting, and routing
  • Database syncs, record updates, and cross-platform data transfers
  • Data transformation, normalization, and enrichment
  • Deduplication, validation, and error flagging

Scheduling & Operations

  • Appointment scheduling, rescheduling, confirmations, and reminders
  • Calendar management and availability checks
  • Job dispatching and assignment based on predefined rules
  • Task routing and status updates across operational tools

Finance & Administration

  • Invoice generation and delivery
  • Payment reminders and overdue follow-ups
  • Expense categorization and receipt processing
  • Document generation from templates (proposals, confirmations, summaries)

Communication (Drafts for Review)

  • Email and message drafts prepared for human review and approval
  • Follow-up sequences triggered by defined events
  • Review request and feedback collection sequences
  • Internal status updates, notifications, and team alerts

Monitoring & Reporting

  • Workflow health monitoring and error alerting
  • Performance dashboards and metric tracking
  • Scheduled report generation and delivery
  • SLA tracking and compliance monitoring

Lead & Customer Management

  • Lead capture from forms, ads, web sources, and inbound channels
  • Initial lead qualification against defined criteria
  • CRM record creation, updates, and lifecycle tagging
  • Review monitoring and reputation tracking

Deterministic Logic & Business Rules

  • Conditional routing based on defined business rules
  • If/then workflows with predetermined outcomes
  • Calculations, conversions, and formula-driven outputs
  • File organization, tagging, archiving, and backup triggers

Exception Process

If a Client needs a task that falls under the HOT categories to be fully automated (not just assisted), the following process applies:

  1. Change Request: The Client submits a written Change Request describing the task, the desired automation scope, and the proposed safeguards.
  2. Exception Assessment: Zeitra.AI leadership reviews the request and evaluates risk, including potential for harm, reversibility, and regulatory implications.
  3. Approval Requirements: Exceptions are only approved when all of the following are in place:

    Written Client sign-off acknowledging the risk
    Human-in-the-loop (HITL) approval gates at critical decision points
    Immutable execution logs for auditability
    A documented rollback plan to revert to human-owned operation if needed

  4. Time-Limited & Auditable: All approved exceptions are time-limited and subject to review. They do not set permanent precedent.

Responsibilities & Enforcement

Client Responsibilities

  • Define HOTs during discovery and approve the HOL before work begins
  • Name human approvers for any workflow that borders HOT territory
  • Handle any required consents, disclosures, or regulatory obligations on their end

Zeitra.AI Responsibilities

  • Enforce the HOL in every build and maintenance engagement
  • Log all workflow executions for auditability
  • Immediately suspend any workflow that crosses HOT boundaries or operates outside the approved scope
  • Conduct quarterly audits for active partnerships; incidents trigger immediate suspension and remediation

Acknowledgment

By engaging Zeitra.AI in any capacity, the Client acknowledges this Policy and agrees to define Human-Owned Tasks during discovery. This Policy applies to all builds, maintenance engagements, and change/improvement work performed by Zeitra.AI.

Contact

Policy questions or exception requests: [email protected]

 Contact