AI Product Terms
Effective Date: 1 July 2026
These AI Product Terms apply to Customer's use of AI features in TurboHelp, including the AI support agent called Turbo, AI-generated replies, summaries, classifications, routing, handoffs, workflow actions, analytics, evaluation tools, and related automation.
1. Relationship to the Terms
These AI Product Terms supplement the TurboHelp Terms of Service. Capitalized terms not defined here have the meanings given in the Terms.
If these AI Product Terms conflict with the Terms, these AI Product Terms control for AI features.
2. Inputs and Outputs
Input and Output are Customer Data.
Customer owns raw Customer Data, including Input and Output, as between Customer and TurboHelp. TurboHelp does not claim ownership of raw Customer Data.
Customer grants TurboHelp the rights described in the Terms and the Data Use and Model Training Addendum to process Input, generate Output, operate AI features, improve the Services, secure the platform, troubleshoot issues, evaluate performance, and provide support.
3. Customer Responsibility for AI Use
Customer is responsible for deciding whether AI features are appropriate for Customer's intended use, business needs, End User expectations, industry, legal obligations, and risk tolerance.
Customer is responsible for:
- Configuring Turbo, knowledge sources, escalation rules, allowed actions, guardrails, and workflows.
- Reviewing Output where appropriate.
- Monitoring AI performance and End User experience.
- Correcting inaccurate or incomplete knowledge base content.
- Deciding when a human should review or take over a support interaction.
- Ensuring that AI-assisted support complies with Customer's policies and applicable law.
TurboHelp is a support automation tool. It is not a substitute for legal, medical, financial, tax, employment, safety, or other professional judgment.
4. AI Output Limitations
AI Outputs may be inaccurate, incomplete, outdated, biased, unsafe, offensive, or unsuitable. Output may misinterpret Input, fail to follow instructions, omit context, hallucinate facts, or produce a response that is not appropriate for a specific End User.
Customer should not rely on Output as the sole basis for decisions with legal, financial, health, employment, housing, credit, insurance, safety, or similarly significant effects.
Customer is responsible for verifying Output before using it where accuracy, safety, compliance, or business impact matters.
5. Notices and Consents
Customer must provide notices and obtain consents for AI-assisted support where required by law, contract, platform policy, or Customer's own commitments.
Customer should disclose to End Users when they are interacting with an AI agent or when AI materially assists support responses if required by applicable law or if disclosure is appropriate for Customer's use case.
Customer is responsible for its privacy policy and End User disclosures, including the disclosures described in the Product Privacy Notice.
6. Knowledge Sources and Grounding
TurboHelp may use Customer-provided knowledge sources, help center articles, product documentation, support history, tickets, chat logs, workflow rules, account context, and other Customer Data to generate, route, summarize, or evaluate responses.
Customer is responsible for the accuracy, legality, completeness, and permissions for knowledge sources and other Customer Data. Outdated, contradictory, or incomplete sources can cause incorrect Output.
TurboHelp may provide tools for testing, previewing, evaluating, or monitoring AI behavior. These tools support Customer's review but do not guarantee that all errors will be detected.
7. AI Testing and Evaluation
TurboHelp will use commercially reasonable efforts to test and evaluate AI features before and after release. Testing may include evaluation for hallucinations, inaccurate answers, unsafe responses, bias, policy-following failures, retrieval quality, escalation behavior, latency, abuse patterns, and other quality or safety issues.
AI systems are probabilistic and may behave differently across contexts, languages, knowledge sources, prompts, End User messages, and product configurations. Testing reduces risk but does not eliminate it.
Customer should conduct its own testing before deploying AI features in production and after material changes to Customer's knowledge sources, workflows, policies, or product.
8. Human Handoff and Escalation
TurboHelp may include handoff, escalation, routing, or confidence-based controls. Customer is responsible for configuring when Turbo should answer, ask clarifying questions, decline, or escalate to a human.
Customer should define escalation rules for sensitive topics, billing issues, account access, complaints, potential legal matters, security incidents, vulnerable users, and any use case where human review is appropriate.
9. Third-Party LLM Providers
TurboHelp may use third-party LLM providers and other AI infrastructure vendors to provide AI functionality.
TurboHelp will contractually restrict third-party LLM providers from using Customer Data to train or improve their own models. Third-party LLM providers may process Input and Output only to provide, secure, monitor, or support AI functionality, subject to applicable subprocessor terms.
Third-party LLM providers are listed or described in the Subprocessors List where applicable.
10. Model Training and Improvement
TurboHelp's use of Customer Data for model training and improvement is governed by the Data Use and Model Training Addendum.
Customer may opt out of use of Customer Data for model training through the mechanism described in that addendum. The opt-out does not prevent use of Usage Data, Aggregated Data, security data, diagnostic data, billing data, or operational analytics needed to provide, secure, analyze, and improve the Services.
Unless Customer has opted out, Customer authorizes TurboHelp to use Customer Data, including End User conversations, Input, Output, feedback, human corrections, successful and unsuccessful resolutions, handoffs, escalation reasons, and workflow traces, to train, fine-tune, evaluate, benchmark, and improve TurboHelp models and AI systems. TurboHelp may use this data to improve retrieval, ranking, summarization, classification, routing, escalation, response generation, safety systems, abuse detection, and quality evaluation.
11. Prohibited AI Uses
Customer must not use AI features to:
- Make decisions with legal or similarly significant effects without appropriate human review.
- Provide medical, legal, financial, tax, mental health, crisis, or emergency advice.
- Generate deceptive, impersonating, manipulative, or fraudulent communications.
- Infer sensitive attributes where prohibited by law.
- Conduct surveillance, profiling, or scoring in a way that violates law or End User rights.
- Create or distribute malware, phishing, spam, or abusive content.
- Process sensitive regulated data prohibited by the No-PHI / No-HIPAA Policy.
12. Alpha and Beta AI Features
AI features marked alpha, beta, preview, pilot, or early access may be less reliable than generally available features. They may change, be suspended, or be discontinued at any time.
Customer should not use alpha or beta AI features for production support or important decisions unless Customer has evaluated the risks and accepts them.
13. No AI Performance Guarantee
TurboHelp may provide metrics such as resolution rate, containment rate, deflection, confidence, quality scores, latency, and customer satisfaction signals. These metrics are estimates and may depend on configuration, data quality, support volume, End User behavior, and measurement settings.
TurboHelp does not guarantee any specific resolution rate, cost reduction, response time, accuracy level, or business outcome unless expressly stated in a signed agreement.