AI and Connector Terms

1. AI-assisted functions

PayCanvas may provide an onboarding agent, generated explanations, policy-drafting assistance, source and mapping suggestions, and connectors that let an authorized user access PayCanvas from a supported AI client. Some functions may be preview or beta features and may change or be withdrawn.

2. External AI clients and native agents

In the connector-first flow, Customer chooses and controls the external AI client. Information disclosed to that client is also governed by the client provider’s terms, privacy practices, account settings, memory, training, and retention controls. The client provider is not our subprocessor merely because Customer independently connects it to PayCanvas. If PayCanvas enables a native agent using a model provider on our behalf, that provider will be identified on our Subprocessors page or in the feature notice.

3. Inputs and permissions

Customer is responsible for the legality, accuracy, quality, and completeness of prompts, source material, instructions, and other inputs. Customer represents that it has the rights and workforce notices or consents needed to provide those inputs and direct their processing. Do not put passwords, API keys, tokens, MFA codes, government identifiers, bank information, health or benefits information, or identifiable employee payroll details into an AI chat. Provider credentials must be entered only through the designated secure PayCanvas form.

4. Outputs require independent review

AI output may be inaccurate, incomplete, stale, biased, or non-unique. It may misunderstand a source system, payroll rule, employee identity, legal requirement, or edge case. Customer must independently verify every material output against authoritative evidence and must not treat generated text as legal, tax, accounting, HR, or employment advice.

Customer may not rely solely on AI output to make a decision that has a legal or material effect on a worker, including compensation, classification, discipline, termination, hiring, scheduling, benefits, or protected-rights decisions. Authorized people must make and document those decisions.

5. Actions and confirmation

An agent may take only the actions exposed by the enabled connector and allowed by the current user’s tenant, role, OAuth grant, MFA level, capability, and service state. Where PayCanvas presents a preview or confirmation, Customer authorizes only the exact action shown. A confirmation to collect sources, save a note, apply a factual resolution, copy selected adjustments, or queue a calculation does not authorize a new policy, payroll approval, export, provider submission, funding, payment, or unrelated action.

6. Policy drafting and ambiguity

The agent may organize Customer statements and evidence into a draft, ask follow-up questions, and identify conflicts. It may not convert an inference, historical outcome, or legacy behavior into approved payroll policy. Unresolved ambiguity must remain visible until an authorized Customer administrator decides it or requests human support. Customer’s approval of the exact policy version remains required.

7. Restricted use

Customer must comply with the Acceptable Use Policy and may not use AI or connector features to evade review controls, access another tenant, infer restricted personal data, make deceptive claims that output is human-reviewed when it is not, develop malware, or create unlawful employment or compensation practices.