Let’s talk

Home / Service Agreements

DEALDOT MEDIA

AI & Site Assistant Schedule

Approved knowledge, human oversight, data boundaries and operational controls for a defined assistant implementation.

Working draft · October 7, 2026. Candidate base terms for review. This document becomes part of an engagement only when identified and accepted in a written client agreement. Complete the applicable placeholders before signing.

1. Application, readiness and definitions

This schedule applies only to an accepted and available AI/site-assistant implementation. DealDot’s public offering remains planning-stage unless readiness is specifically confirmed. “Assistant” means the configured automated interface. “Knowledge Content” means approved source material. “Output” means generated or selected responses. “Escalation” means the agreed handoff to a person or approved destination.

The SOW specifies the provider, functions, channels, knowledge sources, integrations, languages if relevant, permitted data, limits and human contact. A draft schedule does not establish a provider partnership, 24-hour staffed support or readiness of every feature.

2. Authorized purpose and limits

Define the intended tasks, such as explaining approved services, gathering job context or guiding a visitor to an inquiry route. Define what the Assistant must decline or escalate. It must not independently inspect hazards, accept jobs, set binding prices, promise availability or enter Client contracts without separately defined lawful authority and controls.

Disclose the automated nature of the interface. Provide an appropriate human route and instructions for situations outside its remit. The Assistant is not emergency dispatch, regulated advice or a replacement for a professional assessment of a property or trade problem.

3. Knowledge and Client responsibilities

Client supplies accurate, permitted and current Knowledge Content, approved business rules and contact destinations. Identify who can update it, how approvals occur and how obsolete information is removed. Approved knowledge reduces some risks but does not guarantee accurate generated answers.

DealDot configures the agreed implementation and tests representative permitted, unsupported and escalation cases. Client reviews the business claims, availability statements and routing behavior before launch. New knowledge, tools or integrations may require new testing and an accepted change.

4. Human oversight, testing and monitoring

Agree $TEST_PLAN, acceptance criteria, unresolved issues, escalation rules and any $MONITORING_SCOPE. Include tests for misunderstanding, out-of-area inquiries, unsafe requests, unavailable integrations and attempts to elicit secrets or override instructions. Record launch authorization separately from concept approval.

Outputs can be inaccurate, incomplete, biased or inconsistent, and similar responses may be produced for other users. Human review is required where the consequence warrants it. Define who reviews logs, how incidents are reported and when the Assistant should be paused. A continuous-review or error-free promise is not implied.

5. Personal data and provider arrangements

Before use, identify data roles, fields, sources, providers, processing locations, retention, access and any legally required notices or consent. Use the data-processing addendum where applicable. Do not send confidential or sensitive records beyond the agreed purpose and verified provider controls.

Model training, provider improvement use and deletion controls must be stated for the selected configuration. Do not assume a provider excludes training or retains no information. Client data is not authorized for unrelated DealDot training; any provider practice affecting the Engagement must be reviewed and addressed before deployment.

6. Security and integrations

Use proportionate access controls and least-privilege tool permissions. Do not embed secrets in public knowledge or trust user instructions as authorization to access private systems. Identify allowed actions, external destinations, rate/usage limits and confirmation requirements for material changes.

Third-party outages, model changes, prompt attacks and incompatible integrations can affect operation. The Engagement should state provider-change handling and any agreed suspension or fallback procedure. No blanket security certification, uptime SLA or guaranteed resistance to every attack is supplied by this draft.

7. Creative output and rights

Input permissions and provider terms must be checked. AI-assisted output may not be unique or eligible for exclusive protection in every context. Review facts, advertising claims, possible rights conflicts and suitability before public or commercial use.

Ownership or licensing of custom configurations, prompts, knowledge libraries, integrations and reusable DealDot tools follows the accepted IP terms. Raw provider access, editable exports and the underlying model are not automatically transferred to Client.

8. Fees, support and exit

Set $PRICE, $PAYMENT_SCHEDULE, $SERVICE_TERM, usage allowances, overage authorization, provider charges and any $DATE. Define updates, monitoring and support in the SOW; no unlimited interactions, support window or guaranteed escalation response is implied.

On ending services, agree interface removal, integration revocation, access handover, permitted exports and data disposition. Stop presenting an inactive assistant as monitored. Provider deletion, backup retention and configuration portability depend on the actual arrangement. A migration, refund or continuing model access requires an express accepted term.