Screening Tenders with AI: The Smart Go/No-Go Decision

Every week, public procurement platforms flood your inbox with dozens of notifications. You open the links, download the tender documentation packages, skim through the tender regulations. And most of the time, you close the file after an hour, concluding: "not for us." That hour, multiplied by three or four tenders you pass on each week, is half a day of work that vanishes with zero return. AI can bring that time down to five minutes per tender.

This article expands on the Go/No-Go screening section from our complete guide on AI and architecture tenders.

The problem: one to two hours lost per rejected tender

The screening process for a diligent architect almost always follows the same pattern. You spot a tender that seems to match your practice. You download the tender documentation package. You open the tender regulations -- a document of 10 to 25 pages, sometimes more. You read through it, trying to locate the decision-critical information: budget, required references, procedure type, deadlines, location.

Except this information is never gathered in one place. The budget appears on page 4, the selection criteria on page 12, the reference requirements on page 8, the specific competencies in an annex you haven't opened yet. To extract the ten or twelve data points that drive your decision, you have to read most of the document.

The result: between one and two hours per tender reviewed. For a firm that screens five tenders a week and shortlists only two, that amounts to three to six hours every week spent solely on projects you won't pursue. Over a month, that's twelve to twenty-four hours -- two to three full working days.

And the worst part is that this time drain pushes many firms to narrow their search. You end up only looking at the "obvious" tenders -- the ones where you already know the client or the programme. You miss interesting opportunities simply because the cost of screening is too high.

What AI extracts from tender regulations for the Go/No-Go decision

An AI assistant configured for tender screening doesn't read the regulations the way you do. It doesn't scan the document linearly, hunting for what matters. It extracts the decision-critical information directly and organises it into a structured summary. Here are the six categories that drive the decision:

The procedure type. Design competition, simplified procurement procedure (MAPA), open tender, restricted adapted procedure, competitive dialogue. Each type implies a radically different level of effort. A competition requires weeks of work with graphic production. A simplified procurement procedure can be handled in a few hours. This information alone is sometimes enough to decide.

The estimated budget. The construction budget or the design services fee, when stated, tells you immediately whether the project matches your firm's scale. A construction budget of 500,000 euros is not the same game as one of 8 million.

Reference requirements. This is often the most clear-cut elimination criterion. The tender regulations may require three references in public buildings receiving large numbers of visitors, two references in renovations over 2,000 m2, or proven experience on listed heritage sites. If you don't have those references, the discussion is over -- unless you form a consortium.

Location and deadlines. Distance between your office and the project site, submission deadline, projected mission timeline. A tender 400 kilometres away with a ten-day submission deadline raises immediate practical questions.

Selection criteria and their weighting. The split between "references and competencies," "team resources," "methodology statement," and "fee proposal" tells you where to focus your effort -- and whether your profile is favoured or penalised by that weighting.

Specific competencies required. Environmental certification, BIM Level 2 proficiency, experience in seismic zones, acoustics or enhanced accessibility expertise. These requirements may be mandatory or simply valued, and the AI makes the distinction.

How AI cross-references this data with your firm's profile

Extraction alone isn't enough. What makes AI-based screening genuinely effective is the automatic comparison between the tender requirements and your firm's profile.

For this, the AI assistant has access to a document describing your firm: headcount, annual revenue, location, references classified by programme type and budget, certifications, team technical skills, and usual geographic range. You fill this profile in once. It then serves for every screening.

From this cross-reference, the AI produces a point-by-point analysis. For each requirement in the tender regulations, it checks whether your firm meets it -- partially or fully. It identifies your strengths -- criteria where your references exceed the minimum requirements -- and your weaknesses -- criteria where you're borderline or below the threshold.

The output isn't a simple "go" or "no-go." It's a reasoned recommendation with a confidence level. For example: "Go recommended -- strong match on collective housing references and team resources. Watch point: BIM Level 2 certification is not required but appears in the tiebreaker criteria. Consider highlighting existing BIM experience in your application."

Or: "No-Go unless consortium -- references in large public-access buildings are eliminatory and absent from your portfolio. If a partnership with a complementary firm is feasible, reconsider."

Real-world example: a 15-page brief analysed in 5 minutes

Let's take a real, anonymised case. A 15-page tender brief for the restructuring and extension of a secondary school in the Paris region.

You upload the PDF to the assistant. In under three minutes, you get the following summary:

• Subject: Restructuring and extension of Collège Victor Hugo -- Base design services mission + partial detailed design
• Procedure: Restricted adapted procedure, application phase then offer phase
• Client: Val-de-Marne County Council
• Construction budget: EUR 8.2M excl. VAT
• Submission deadline: 21 days (4 May 2026)
• Required references: 3 references in public facilities in the last 5 years, including at least 1 large public-access building, construction budget over EUR 3M excl. VAT
• Selection criteria: References and competencies (40%), team and resources (25%), project understanding and methodology (35%)
• Specific competencies: Experience in occupied-site works preferred, environmental approach valued, awareness of school-schedule constraints mentioned
• Team composition: Consortium with structural and MEP engineering consultants required

Then the recommendation:

No-Go unless consortium. Your firm has 4 qualifying references in public facilities (Créteil media library, Vitry school complex, Montreuil cultural centre, Vincennes sports hub). However, none qualifies as a large public-access building -- your projects fall into smaller classification categories. This criterion is eliminatory. A consortium with a firm that holds this reference would make the application viable. Note: the strong weighting on methodology (35%) favours you given your experience with occupied-site works.

In five minutes, you have an informed decision, the precise reasons behind it, and even an action path if you still want to compete. Compare that with the hour you'd have spent reading the document end to end to reach the same conclusion.

What AI can't decide for you

The screening summary and recommendation are decision-support tools, not the decision itself. Several factors fall outside automated analysis and rely on your judgement as an architect and firm leader.

Strategic alliances. The AI can flag a missing reference and suggest a consortium. But it doesn't know that you have a partnership under discussion with a particular firm, that an engineering consultant proposed a collaboration last week, or that a colleague is also eyeing this tender and a joint bid would make sense.

Political context. You know the local mayor just changed and the new administration has different architectural preferences. You know a certain social housing developer has a complicated history with out-of-town firms. You know this competition is probably earmarked for a local architect. No AI has access to this information.

The desire to break into a new market. The AI will say "No-Go" if you lack healthcare references and the tender is for a care home. But maybe that's precisely the programme type you want to develop, and you're willing to invest time on a less competitive application to start building your portfolio in that area. That strategic decision is yours.

Your firm's current workload. A tender flagged "Go" by the AI might land at the worst possible time -- in the middle of a planning application deadline, during a key team member's holiday. Only you know your firm's operational context.

AI handles the extraction and comparison. The decision remains human, strategic, and informed by factors only the architect knows.

The real gain: from 1-2 hours to 5 minutes per discarded tender

Let's do the maths. A five-person firm that reviews five tenders a week and shortlists two on average.

Without AI: 5 tenders x 1.5 hours of reading = 7.5 hours per week on screening. Over a month, that's 30 hours. On the three discarded tenders alone, 4.5 hours are purely wasted each week.

With a configured AI assistant: 5 tenders x 5 minutes of analysis = 25 minutes per week. Even adding 10 minutes of review per summary, you're at 75 minutes per week. Over a month, that's 5 hours instead of 30.

The gain: 25 hours per month -- more than three working days recovered. Days you can spend writing the tenders you actually pursue, working on current projects, or casting a wider net that opens up new opportunities.

And this is just screening. To discover how AI saves you time on the other stages -- analysing tender regulations, writing your application, building the technical offer, managing admin documents -- see our complete guide.

The financial case is straightforward: if those three recovered days let you respond to one additional tender per month, and your win rate is 15%, you land one to two extra commissions per year. At the average fee for a design services contract, the return on investment is immediate.

Super Archi includes a ready-to-use Go/No-Go screening assistant, already configured with the vocabulary of architecture public procurement. You fill in your firm's profile once, then feed it your tender briefs as they come in. Five minutes per tender, a reasoned recommendation, and you move on.

👉 Discover Super Archi