AI Usage Disclosure

1. Where AI Is Used

TenderBridge routes different tasks to different AI providers, chosen for what each is strongest at:

• Tender and grant match scoring — OpenAI models estimate how well an opportunity fits your organization profile and explain the reasoning in plain language.
• AI Tender Feed recommendations and proposal drafting — OpenAI models assist with bid strategy suggestions and draft proposal content in the Bid Assistant.
• Business Assistant chat — OpenAI models answer questions about your pipeline and how to use the platform.
• Document and website extraction — Anthropic (Claude) models read PDFs, Word documents, and web pages to extract structured tender/grant fields for admin review before publishing.
• Compliance and eligibility review — Anthropic models cross-check extracted requirements against standard eligibility criteria.
• Advanced Search profile extraction — Anthropic models turn a company website, description, and your answers to three questions into a structured search profile (sectors, countries, funder/buyer types, typical value range) used to score every tender and grant.

2. No Fully-Automated Decisions With Legal or Similarly Significant Effect

AI on TenderBridge is advisory. It surfaces information, suggests matches, and drafts content — it does not decide whether you win a tender or grant, and it does not make any decision about you that has a legal or similarly significant effect. Buyers, funders, and evaluation committees — not TenderBridge's AI — make award decisions. Match scores are estimates to help you prioritize your time, not guarantees or professional judgments.

3. Human Oversight and Your Responsibility

Tenders and grants ingested via AI extraction are reviewed by a human administrator before being published to the platform. Content drafted by the Bid Assistant is a starting point, not a final submission — you are responsible for reviewing, verifying, and editing any AI-generated content before relying on it or submitting it to a third party. If you believe a match score, extracted field, or AI response is wrong, you can disregard it and rely on the original source document or listing, which we link to wherever available.

4. Data Sent to AI Providers

Depending on the feature, the data sent to our AI providers may include: your organization profile (sectors, geography, contract value range), tender or grant document text or the content of a webpage you or an admin submitted, the description and answers you provide to Advanced Search, and your messages to the Business Assistant. We do not send your payment or billing details to AI providers. Under our current agreements, OpenAI and Anthropic do not use this data to train their general-purpose models.

5. Reliability and Rate Limiting

AI requests are automatically retried on transient failures and, for supported tasks, can fail over from one provider to the other to keep the feature available. Every request is logged (task type, provider, token usage, and cost) for reliability monitoring — administrators can review this in the AI Usage dashboard. Feature-specific rate limits apply to prevent abuse and manage cost; you'll see a clear message if you hit one.

6. Limitations

AI models can make mistakes: they may misread a document, miss context, or produce a plausible-sounding but incorrect answer. Match scores are heuristic estimates based on the information available and should be treated as a starting point for your own judgment, not a substitute for it. Always verify extracted deadlines, values, and eligibility requirements against the original tender or grant document before acting on them.

7. Questions or Objections

If you have questions about how AI is used on TenderBridge, or want to raise a concern about an AI-assisted decision that affected you, contact us at admin@growthinvest.org.