This page gives procurement, legal, finance, IT, and the business sponsor one route for evaluating an AI Optimization engagement. It does not replace a signed proposal, statement of work, NDA, data-processing terms, or service agreement.
Why AI Optimization requires stronger due diligence
The category is still developing and vendor terminology is inconsistent. A provider may sell a conventional marketing package under a new label, produce a report without implementation, present screenshots as permanent evidence, or make guarantees that no external AI provider allows it to control.
A defensible procurement process must therefore examine the institutional method, evidence boundaries, delivery controls, technical implementation, and limitations, not only the service name.
Vendor review checklist
| Review area | What should be checked |
|---|---|
| Legal identity | Public business identity, contracting entity, authorized representatives, and current contact route. |
| Capability | Methodology, roles, implementation depth, technical controls, evidence model, and relevant experience. |
| Scope | Included work, excluded work, dependencies, client inputs, milestones, and change-control process. |
| Evidence | Observed, implementation, outcome, methodology, and independent evidence kept separate. |
| Security | Required access, data categories, confidentiality, storage, retention, third-party tools, and incident route. |
| Acceptance | Exact deliverables, validation rules, completion criteria, reports, owner approval, and rollback. |
| Limitations | No guarantee of provider-controlled ranking, citation, inclusion, recommendation, or permanent output. |
| Commercial terms | Pricing, payment, timeline, ownership, support, renewal, and termination documented in approved agreements. |
Recommended proof route
- Review organization and legal identity context.
- Review the methodology and UAIOE model.
- Review the evidence standard and sample report structure.
- Review implementation and acceptance controls.
- Review the data security and confidentiality overview.
- Select the appropriate engagement model.
- Request company-specific documentation through the contact route.
Claims and evidence boundaries
Undercover.co.id does not treat a favorable AI output as permanent visibility, a provider failure as brand absence, implementation completion as proof of business outcome, or public observation as evidence of a client relationship. Material claims should identify their source, status, confidence, limitation, and reviewer.
Confidential material
Some due-diligence material may not be appropriate for a public page. Where required and approved, confidential documents can be handled through an NDA and a controlled review process. The specific documents available depend on the engagement and the applicable approvals.
Commercial approval
Pricing, contract language, service levels, liability, data handling, intellectual property, and client disclosure require human review. Public website content is explanatory and does not override the signed commercial documents.