Analyzing Tender Regulations with AI

This article is part of the complete guide: AI and architecture tenders.

You've decided to respond. The Go/No-Go is behind you, the tender matches your firm. Now the real work begins. And it all starts from the same document: the tender regulations.

The tender brief is the rulebook. Everything is in there -- the evaluation criteria, the required documents, the deadlines, the specific constraints. But everything is also scattered, buried in dense legal language, spread between the body text and the annexes, sometimes amended by an addendum published three days before the deadline. A misreading of the tender regulations can't be fixed later. If you miss a weighting criterion or forget a required document, your application is eliminated or penalised -- regardless of the quality of your references or the relevance of your approach.

This is precisely where AI makes a measurable difference.

What tender regulations contain -- and why they're painful to read

A typical set of tender regulations for an architecture public tender runs 15 to 30 pages. It covers the subject of the contract, participation conditions, selection criteria, submission requirements, required documents, the timeline, and a set of legal clauses -- some of which contain operationally critical information.

The problem isn't the length. The problem is the structure. The information you need is never grouped in a logical place. The main weighting criterion is mentioned on page 8, but a decisive sub-criterion appears on page 22, in a paragraph that begins with "as a supplementary consideration." The list of required documents is on page 12, but an annex on page 27 specifies that a mandatory reference template is imposed -- a detail that completely changes your formatting work. The site visit is mentioned in a footnote. The consortium requirements are detailed in the contract terms, not in the tender brief itself.

The result: to extract the genuinely useful information from a 25-page brief, an experienced architect spends one to two hours. Reading, highlighting, note-taking, going back, cross-referencing with the contract terms and the programme document. And when you're handling three tenders in parallel, information bleeds together. Criteria from one get confused with constraints from another. Details slip through the cracks.

This isn't a competence problem. It's a volume and format problem. A 25-page legal document isn't designed to be read like a project brief. It's designed to be legally exhaustive -- not to make your life as an applicant easier.

What AI extracts from tender regulations

An AI assistant configured for architecture public procurement reads the entire brief (and ideally the contract terms and programme document as well) and produces a structured extraction. Here's what it identifies and organises:

Selection criteria with their exact weighting. This is the most strategically important information in the document. "Technical value: 60% (of which methodology 30%, references 20%, team resources 10%) -- Fee proposal: 40%." These numbers dictate how you allocate your time and energy across each part of your response. The AI spots them even when they're scattered across multiple articles.

Required documents, with their specifications. Not just "a cover letter and a DC2" (DC2 is a standard French procurement form declaring the candidate's financial and professional capacity), but the detail that matters: maximum number of references (often 5), imposed format (template provided by the client or free-form), page limits for the methodology statement, required file formats (PDF, spreadsheet for the bill of quantities).

Specific constraints. Mandatory site visit (with date and registration details). Variants allowed or prohibited. Required consortium with specific competencies (landscape architect, structural engineer, cost consultant). Whether the previous competition winner retains any priority. Required environmental certification levels.

The full timeline. Application or offer submission deadline (date and time). Site visit date. Estimated review period. Estimated notification date. Offer validity period. In a two-phase procedure, the dates for each phase.

Key programme elements. Projected floor area, budget envelope, number of lots, project phasing, site constraints (occupied building during works, asbestos presence, regulated zone).

The classic mistake: the criterion hidden in an annex

Every architect who has responded to enough tenders knows this situation. You've read the brief, noted the criteria, prepared your response. Then during the final review, a colleague spots a paragraph on page 26: "Applicants must demonstrate at least two references in the renovation of listed heritage buildings within the past five years." Or: "Sub-criterion 2.3, detailed in Annex C, will assess the applicant's ability to propose a BIM Level 2 approach."

This kind of information is a game-changer. If you don't have the listed-building references, you're wasting your time applying. If you haven't anticipated the BIM sub-criterion, your methodology statement will be off-target on 30% of the technical score.

The fundamental advantage AI has here is simple: it doesn't get tired, and it doesn't skim. An architect reading their third brief of the week at 6pm on a Friday will involuntarily speed through paragraphs of legal jargon. The AI processes page 28 with the same attention as page 3. It catches decisive information wherever it sits -- in the body of the brief, in an annex, in the contract terms, in an addendum posted on the procurement platform.

And when you feed it the entire tender documentation package rather than just the brief, it can cross-reference between documents. A deadline mentioned in the brief but clarified in the contract terms. A technical constraint in the programme document that has implications for the selection criteria. This multi-document cross-referencing is something architects rarely do for lack of time -- and that AI does systematically.

How AI structures the extraction: summary sheet vs narrative summary

There's a fundamental difference between a summary and a structured extraction. A summary rephrases the brief into shorter prose -- that's what a generic ChatGPT produces if you ask it to "summarise this document." The result is shorter than the original, but it's still continuous text you have to search through.

A structured extraction, on the other hand, reorganises information by category and priority. It's a working document, not a text to read. Here's the concrete difference.

A properly configured assistant for architecture tenders produces a summary sheet that clearly separates:

• A criteria table with weighting, sub-criteria, and references to the relevant articles
• A document checklist with formats, limits, and imposed templates
• A timeline with all dates and deadlines
• A list of watch points: anything unusual, restrictive, or potentially eliminatory

This format isn't an aesthetic choice. It's an operational one. Each part of this sheet feeds directly into a stage of your work. The criteria table guides the structure of your methodology statement. The document checklist becomes your pre-submission control list. The timeline goes into your planning. The watch points steer your strategic choices.

The output format: your working document for the entire response

Concretely, here's what you get after running the tender brief through the AI assistant:

Selection criteria table

Required documents checklist

• Cover letter (DC1 form -- standard French procurement form identifying the candidate)
• Candidate declaration (DC2 form -- standard French procurement form on financial/professional capacity)
• Current professional indemnity insurance certificate
• Maximum 5 references, using imposed template (Annex 3)
• Methodology statement -- 10 pages maximum, A4 format
• Projected timeline
• Lump-sum fee breakdown (using provided response template)

Timeline

• Site visit: 15 April 2026, 10am -- mandatory registration before 12 April
• Submission deadline: 30 April 2026, 12:00pm
• Estimated notification: July 2026

Watch points

• Site visit is mandatory -- absence is eliminatory
• Consortium with an RGE-certified MEP engineering consultant required (RGE is a French energy-efficiency qualification for construction professionals)
• The "environmental approach" sub-criterion (Annex C) accounts for 15% of the technical score -- not visible in Article 8.1 of the brief
• Variants not permitted
• Offer validity period: 180 days (instead of the usual 90 days)

This document fits on one to two pages. It replaces the 25-page brief as your daily reference throughout the preparation of your response. You only open the original PDF to verify a specific point -- and you know exactly where to look thanks to the article references.

A working document that feeds everything downstream

The value of this structured extraction goes beyond simply reading the brief. It becomes the foundation of your entire response:

For the application, the criteria table tells you where to put the effort. If references account for 20% and team resources for 10%, you devote twice as much content to your references as to your org chart. The checklist ensures you don't miss any required document.

For the technical offer, the detailed sub-criteria become the outline of your methodology statement. If the brief mentions three sub-criteria (programme understanding, project management approach, environmental strategy), your statement has three sections -- in that order, with volume proportional to the weighting.

For the admin package, the document checklist is your final quality control before submission. Each line ticked off is one less risk of elimination.

For your internal planning, the extracted timeline lets you set up your reverse schedule immediately: site visit date, submission deadline, and therefore intermediate dates for drafting and review.

As explained in the complete guide, AI doesn't replace your strategic judgement. But it gives you the elements to exercise that judgement on a clear and complete basis, instead of working from a hasty reading of a 30-page PDF.

Real-world example: before and after AI extraction

Before -- what you read in the brief (raw excerpt):

"Applications will be assessed against the following criteria, in accordance with the provisions of Article R. 2142-17 of the French Public Procurement Code: 1) Professional guarantees; 2) Technical and financial capacity; 3) Professional references. The weighting of criteria is detailed in Article 8.1 of this document. Applicants are informed that the contracting authority reserves the right to verify the certificates and attestations produced, in accordance with Article R. 2143-6. Furthermore, applicants are invited to refer to Annex C for the detailed sub-criteria relating to technical value. A site visit is recommended; arrangements are specified in Article 4.3. Applicants must produce all documents listed in Article 6, it being understood that any incomplete application may be rejected."

Two paragraphs later, in another article: "Applicants' attention is drawn to the fact that the site visit is mandatory." This contradicts the "recommended" wording in the earlier extract -- the term was corrected by addendum, but the original passage was never updated.

After -- what the AI assistant produces:

Site visit: MANDATORY (Art. 4.3, corrected by Addendum No. 1 dated 03/04/2026 -- note: the body of the brief states "recommended" but the addendum makes the visit mandatory and eliminatory).

The AI identified the contradiction, flagged the addendum, and retained the version in force. An architect in a hurry who read only the body text without the addendum would have missed the visit -- and their application would have been eliminated on a purely procedural ground.

This is exactly the kind of gain that AI analysis of tender regulations delivers. Not a comfort gain -- a reliability gain. Fewer errors, fewer oversights, and a solid starting point for every subsequent stage of your response.

Super Archi includes a dedicated tender regulations analysis assistant. You upload the brief (and the rest of the tender documentation package if you wish), and within minutes you get a structured summary sheet, ready to use for preparing your application and technical offer.

👉 Discover Super Archi