Responsible AI.

How AI Works Inside Sentiva

Sentiva uses trusted third-party AI models through secure providers, combined with Sentiva's HR-specific workflows, prompts, rubrics, guardrails, and evaluation logic.

These AI-assisted workflows support features such as candidate evaluation, structured interview analysis, resume review, panel feedback summarization, assessment insights, report generation, and recruiter-facing recommendations.

Sentiva does not make final employment decisions. AI outputs are designed to support human review, not replace it. Hiring, employment, promotion, termination, compensation, and other people-related decisions remain with the customer's authorized users.

For supported hiring evaluation workflows, Sentiva maintains reviewable evidence where applicable. This may include the source information considered, why that information was relevant, how it mapped to the applicable rubric, and the rationale behind the resulting score, summary, or evaluation output.

Our Responsible AI Principles

1. AI supports decisions. People make them.

AI-generated outputs inside Sentiva are recommendations, summaries, analyses, or decision-support signals. They are not final decisions.

Sentiva does not automatically hire, reject, advance, terminate, promote, or disqualify individuals on its own. Final decisions remain with the customer's authorized users.

Customers are responsible for applying appropriate human review, considering the full context, and complying with employment, labor, anti-discrimination, privacy, AI, and human rights laws that apply to their organization and region.

2. AI is reviewable.

For AI-assisted workflows that affect candidate or employee evaluation, Sentiva aims to provide supporting evidence where applicable. This may include relevant resume excerpts, transcript references, rubric mapping, score justification, and evaluation rationale.

For other AI-assisted features, such as summarization or drafting assistance, Sentiva may provide generated outputs without a full evidence chain. Users should review these outputs before relying on them.

3. AI is auditable where it matters most.

For supported evaluation workflows, Sentiva retains relevant AI outputs, evidence references, configuration details, and audit events according to the applicable customer agreement and retention settings.

This information may not always be directly exportable from the product interface, but customers can request export or review support where required for legal, regulatory, internal audit, or governance purposes.

4. AI is configurable.

Sensitive AI-assisted features, such as resume scoring and AI interviews, can be enabled, disabled, or scoped based on the customer's workflow and product configuration.

Not every Sentiva workflow requires AI-assisted evaluation. Customers are responsible for configuring Sentiva appropriately for their policies, regulatory obligations, and hiring or HR practices.

5. AI is bounded by permission and context.

Lumi, Sentiva's AI interview agent, and other AI-assisted workflows only access data required for the specific workflow being performed.

Sentiva does not share Customer Data across customer tenants by default. Customer Data from one tenant is not used to train models for another customer.

6. AI is designed for sensitive HR use cases.

Sentiva does not currently train its own AI models using Customer Data. Instead, Sentiva uses third-party AI models through controlled HR-specific workflows, prompts, rubrics, sanitization steps, and evaluation logic.

Where possible, Sentiva limits the personal information sent to AI providers. Information that is not relevant to the workflow, such as names, gender, age, or other unnecessary identifiers, may be excluded or sanitized before processing to reduce bias risk and unnecessary data exposure.

Human Oversight and Customer Responsibility

Sentiva is a decision-support system. It is not a replacement for recruiter, HR, manager, legal, or compliance judgment.

Customers are responsible for:

• reviewing AI-assisted outputs before relying on them;
• ensuring appropriate human oversight;
• deciding whether and how AI-assisted features should be used in their jurisdiction;
• providing required notices and obtaining required consents or lawful grounds;
• responding to candidate or employee requests;
• ensuring their use of Sentiva complies with applicable employment, labor, anti-discrimination, privacy, AI, data protection, and human rights laws; and
• making all final hiring, employment, promotion, termination, compensation, disciplinary, compliance, and business decisions.

Customers should not use AI-assisted outputs as the sole basis for employment or similarly significant decisions.

Data Ownership and Processing

Customers own their Customer Data.

For HR data that customers upload or generate inside Sentiva, Sentiva generally acts as a data processor or service provider. The customer remains the controller or business for candidate, employee, and HR data processed on its behalf.

For certain website, account, billing, support, telemetry, marketing, or future candidate-facing services, Sentiva may act as an independent controller or business as described in our Privacy Policy and applicable terms.

Our role and obligations are governed by our customer agreements, Terms of Service, Privacy Policy, and Data Processing Addendum where applicable.

In practice:

• Customer Data is owned by the customer.
• Sentiva processes Customer Data according to documented customer instructions and applicable agreements.
• Sentiva does not sell personal information for money.
• Customer Data is not used to train external or third-party foundation models.
• Customer Data is not used to train models for other customers.
• Sentiva's AI provider arrangements are designed to prevent Customer Data submitted through Sentiva from being used by those providers for model training or model improvement.
• Any use of aggregated usage analytics to improve Sentiva is handled in accordance with our Privacy Policy, customer agreements, and applicable law.
• Customers can request export of supported Customer Data, subject to the applicable customer agreement and product capabilities.

Model Training and Product Improvement

Customer Data is not used to train external or third-party foundation models.

Sentiva does not currently train its own AI models using Customer Data. If Sentiva develops internal models in the future, Customer Data will only be used where permitted by the applicable customer agreement, law, and documented privacy safeguards.

Where appropriate, Sentiva may use aggregated, de-identified, pseudonymized, or otherwise protected data to improve product quality, workflow design, prompts, rubrics, evaluation logic, analytics, reliability, and security.

Sentiva also improves AI-assisted workflows through HR-specific prompts, rubric design, internal evaluation sets, quality reviews, customer feedback, product analytics, and workflow testing, subject to our privacy commitments and customer agreements.

Candidate and Employee Rights

Where Sentiva is used in hiring or HR processes, individuals may have rights under privacy, employment, AI, accessibility, and anti-discrimination laws.

For candidates interacting with Sentiva-powered hiring workflows:

• Sentiva is designed to inform candidates where AI-assisted evaluation is part of the workflow.
• Before a Lumi interview, candidates are informed that the interview is AI-conducted.
• Candidates using Lumi may request an alternative evaluation path through the hiring organization.
• Declining or requesting an alternative to a Lumi interview does not automatically disqualify a candidate within Sentiva.
• The hiring organization is responsible for deciding how to handle alternative evaluation requests.
• Data subject rights requests should generally be made through the hiring organization.
• Where Sentiva receives a rights request directly, we will support the relevant customer in accordance with applicable law and our agreements.

For employees of Sentiva customers:

• Employee data is primarily governed by the employer's data policies and the customer's agreement with Sentiva.
• Data subject rights requests should generally be made through the customer's HR or privacy team.
• Where required, Sentiva supports customers in responding to verified rights requests within applicable legal timeframes.

Candidate-Facing Transparency

Candidate-facing AI disclosures appear at relevant points in the hiring workflow where AI-assisted evaluation occurs.

This may include:

• on the application form, where AI-assisted screening is configured by the hiring organization;
• before a Lumi interview, where the candidate is informed that the interview is AI-conducted and given the option to request an alternative path; and
• in candidate-facing communications, where applicable.

Disclosures explain that AI may be part of the evaluation, what role it plays, and that final decisions are made by humans.

The specific wording may be configurable by the hiring organization to fit its tone, language, and regional regulatory requirements.

What Sentiva's AI-Assisted Workflows May Do

Depending on product configuration, Sentiva's AI-assisted workflows may support:

• candidate scoring based on resumes, transcripts, structured questions, or rubrics;
• structured AI interview analysis;
• interview transcript summarization;
• panel feedback summarization;
• recruiter-facing recommendations;
• review flags for human consideration;
• report generation;
• evidence mapping;
• assessment insights; and
• pattern surfacing within supported HR workflows.

What Sentiva's AI-Assisted Workflows Do Not Do

Sentiva's AI-assisted workflows do not:

• make final employment decisions;
• automatically hire, reject, advance, terminate, promote, or disqualify individuals on their own;
• operate outside the customer's configured permission scope;
• intentionally access data outside the customer's tenant;
• use video appearance, facial features, facial expressions, ethnicity, gender, age, or similar characteristics to determine candidate evaluation scores;
• score candidates based on vague or unsupported concepts such as “culture fit”; or
• replace the customer's obligation to review, validate, and lawfully use outputs.

Where Sentiva provides behavioral or psychometric assessments, those assessments use structured assessment methods and scoring approaches. They are not used by AI to infer personality from video, appearance, facial expression, or unrelated personal characteristics.

AI Interviews and Video

AI interviews may be video recorded where enabled by the customer and disclosed to the candidate.

The evaluation itself is based on the candidate's responses and the applicable rubric, not facial analysis, visual appearance, or physical characteristics.

Sentiva may separately review interview conditions for integrity or proctoring-related signals, such as whether the camera was not visible, the image was too dark, or more than one person appeared on screen. These signals are treated separately from the evaluation score and are flagged for human review where relevant.

Fairness, Bias, and Accessibility

Sentiva is designed to reduce unnecessary subjectivity in hiring and HR workflows by using structured prompts, role-relevant rubrics, consistent evaluation criteria, and reviewable evidence.

Sentiva's fairness practices may include:

• pre-deployment testing of supported workflows and rubrics against internal evaluation sets;
• review of scoring behavior and output consistency;
• monitoring for statistical patterns that warrant further review;
• sanitization or exclusion of unnecessary personal identifiers where the workflow does not require them;
• rubric-based scoring designed around job-relevant criteria; and
• human review of outputs before decisions are made.

Sentiva does not use candidate names, gender, age, ethnicity, video appearance, facial features, or similar characteristics as scoring criteria for candidate evaluation.

Bias testing methodology summaries and available evaluation documentation may be provided to enterprise customers under NDA where applicable.

Sentiva also recognizes that AI interviews may not work equally well for everyone, including people with speech impairments, hearing impairments, neurodivergent communication styles, anxiety-related barriers, language differences, or accessibility needs. Candidates may request an alternative path through the hiring organization.

Model Governance and Change Management

Sentiva uses third-party AI models, which may be updated by the underlying providers over time.

When a material model or workflow update may affect supported evaluation workflows, Sentiva may:

• test the updated model or workflow against internal evaluation sets before broader rollout;
• document model or workflow changes where relevant;
• monitor output quality and behavior;
• notify enterprise customers of material changes where required by agreement; and
• work with customers that have model-version consistency requirements where commercially and technically feasible.

Enterprise customers may request available AI governance documentation, including summaries of supported use cases, limitations, evaluation practices, workflow controls, and change management practices.

Regulatory Alignment

Sentiva is designed to support customers in meeting applicable privacy, security, HR, and AI governance obligations. However, customers remain responsible for ensuring their use of Sentiva complies with the laws that apply to their organization, candidates, employees, and operating regions.

Sentiva monitors evolving AI and employment-related requirements, including the EU AI Act, U.S. state and local rules relating to automated employment decision tools, GDPR, UK GDPR, CCPA/CPRA, Sri Lanka PDPA, and other applicable data protection and AI governance frameworks.

Customers are responsible for determining whether their use of Sentiva requires notices, consents, human review, impact assessments, bias audits, accessibility accommodations, worker consultation, or other compliance steps.

Where cross-border transfers or additional contractual safeguards are required, Sentiva will work with customers to assess appropriate mechanisms under the applicable agreement.

Sentiva applies technical and organizational measures such as access controls, encryption, tenant separation, data minimization, logging, and workflow-level permissioning to help protect Customer Data.

Privacy and data protection inquiries can be raised through the contact details provided in our Privacy Policy.

Known Limitations of AI-Assisted Workflows

Sentiva believes transparency about AI limitations is part of responsible deployment.

Customers and users should understand that:

• AI-assisted summarization may miss nuance or subtle signals a human reviewer would catch.
• AI-assisted scoring is one input among many and should not be used as a final decision on its own.
• AI-assisted evaluation may be less consistent across communication styles, accents, language backgrounds, or non-standard communication patterns.
• AI systems may produce incorrect, incomplete, biased, or unexpected outputs.
• AI-assisted workflows depend on the quality and relevance of the data provided.
• Sentiva is not designed to evaluate legal compliance, medical or disability accommodations, security clearances, or specialized assessments outside its supported HR workflow scope.

We disclose these limitations because responsible AI use requires realistic expectations, human review, and careful customer configuration.

What Happens if Something Goes Wrong

Sentiva maintains incident response procedures for security and privacy events.

In the event of a security or privacy incident affecting Customer Data:

• where required, Sentiva will notify affected customers without undue delay so they can assess and meet their own legal and regulatory obligations;
• Sentiva will work with customers to assess impact, contain the issue, and support remediation; and
• affected individuals will be notified by the appropriate party where legally required.

Concerns about AI outputs or platform behavior can be raised through Sentiva support or the designated privacy/contact channel.

Sentiva will investigate concerns raised by customers and, where appropriate, support the customer in reviewing the relevant workflow, evidence, configuration, or output.

For Evaluation Teams

If your governance, security, legal, HR, or compliance team needs to evaluate Sentiva in detail, we can support your review process.

Available on request where applicable:

• Data Processing Addendum;
• security documentation;
• AI governance documentation;
• supported workflow documentation;
• data processing and subprocessor information;
• available audit, review, or testing summaries; and
• customer-specific contractual documentation.

Public resources:

• Privacy Policy;
• Terms of Service;
• Cookie Policy;
• Acceptable Use Policy; and
• Accessibility Statement.

Privacy and data protection contact details are available in our Privacy Policy.