AI tender analysis

When you upload a tender document, the text is extracted and passed to a language model with a defined extraction task. The model is not asked for an opinion on whether you should bid. It is asked to identify and return specific fields that the document contains:

• tender title, reference number and contracting authority;
• estimated contract value, duration and extension options;
• eligibility requirements and mandatory documents;
• evaluation criteria and their published weightings;
• key dates, including clarification windows and the submission deadline;
• risks and compliance items stated in the pack.

The result is a structured record of the document, not a rewritten summary of it.

This distinction matters. Everything downstream of extraction is deterministic application logic, not AI judgement.

Company fit

Requirements are compared against your saved company profile by fixed rules — turnover thresholds, insurance levels, certification matching — and marked met, review, or not met. The same profile and the same tender always produce the same result.

Risk classification

Risks are classified from the evidence in the document. An overall high risk position can only arise where at least one individual risk has evidence supporting it. Where the pack simply does not say enough, risks are recorded as not stated rather than assumed severe.

Decision Score and recommendation

The 0–100 Decision Score and the GO, REVIEW or NO-GO recommendation are calculated by rules you can reason about, including an absolute override where a mandatory requirement is not met. No model is asked to choose the verdict.

Language models make mistakes, and tender packs make those mistakes more likely:

• Scanned or image-based PDFs extract poorly, and text that cannot be read cannot be analysed.
• Multi-document packs mean a single upload may not contain every requirement; terms, pricing schedules and annexes are often issued separately.
• Figures and dates can be misread. TenderMet flags dates that look chronologically implausible for you to check, rather than correcting them silently.
• Ambiguous drafting — where a requirement is genuinely unclear, the correct output is review, not a confident answer.
• Addenda and clarifications issued after you downloaded the pack will not be reflected in an earlier analysis.

TenderMet's output is decision support. It helps you see what a document contains and how it compares with your company profile, so a human can make a faster and better-informed judgement. It is not legal advice, procurement advice, financial advice or a compliance certification, and it does not guarantee any tender outcome.

Before submitting anything, verify every requirement, figure and date against the original documentation on the authority's portal. Where a point is material and unclear, raise a clarification question with the authority within the published window.

Uploaded documents are processed to produce your assessment, and assessments are saved to your own account so you can reopen them without re-analysing. Access is restricted to your account. Full detail is set out in our Privacy Policy, and our Terms of Service set out the basis on which the analysis is provided.

Related reading: tender analysis software and UK tender evaluation.