Construction tender comparison with ChatGPT: method and checks
Construction tender comparison with ChatGPT: the most common VAT discrepancy, equal treatment in public contracts and the method.
By Educasium
Install the ChatGPT skill: Construction tender comparison →
Two contractors respond to the same tender for renovating a bathroom. Contractor A quotes €12,000 including installation; contractor B quotes €10,500 but does not specify whether installation is included. An architect in a hurry might be tempted to rank B first based on the displayed amount alone. A closer look reveals a second, quieter gap: A applied 10% VAT, B applied 20% VAT, on what looks like the same scope of work — a difference that alone can amount to several hundred euros and says nothing about the quality of the bid, only about how each contractor understood the applicable tax regime.
This article details why a bid comparison must never stop at the displayed amounts, the difference between a public contract subject to equal treatment of bidders and a private contract where that rule does not apply, the most common VAT discrepancy between apparently comparable bids, and the method for building a reliable comparison with ChatGPT.
Summary
- Why a displayed amount is never enough to compare two bids
- Public or private contract: two different regimes for ranking bids
- The VAT discrepancy hiding behind apparently comparable bids
- Comparing bids with ChatGPT, step by step
- When this comparison is not enough
- Training to make bid analysis more reliable
- Frequently asked questions
Why a displayed amount is never enough to compare two bids
A total amount, on its own, says nothing about a bid's actual scope: options included or not, discounts, taxes, installation method, warranties offered. Two contractors can respond to the same CCTP with different assumptions about what their price covers, with that difference showing up nowhere except in the line-item detail.
Mapping every item to a common item list, without rewriting the original bids
The first task of any comparison is to map each bid's items to a common item list, while keeping the original bids intact as a reference. An item that contractor A describes in one line and contractor B details across three lines must be explicitly matched, never silently merged at the risk of losing part of the real scope.
Public or private contract: two different regimes for ranking bids
How bids are ranked depends directly on the client's status, a point many comparisons ignore by applying the same method regardless of who the buyer is.
| Client status | Rules governing ranking | What this requires in practice |
|---|---|---|
| Public buyer (public body, public establishment) | Free access, equal treatment of bidders and transparency of procedures | Objective, weighted criteria announced in advance; no adapting the scoring method after bids are received |
| Private client (individual, company, developer) | Contractual freedom, no legal framework imposing a ranking method | The client may choose their own criteria freely, but must document them to stay consistent across bids |
The principle of equal treatment, specific to public contracts
For a public buyer, Article L3 of the French public procurement code sets out free access to public procurement, equal treatment of bidders and transparency of procedures. This principle in particular forbids adapting the scoring method after bids have been received, or imposing requirements only one already-identified bidder could meet. This framework does not apply to a private contract, where the client retains much wider contractual freedom — but where consistent criteria remain essential for the comparison to stay defensible.
The VAT discrepancy hiding behind apparently comparable bids
A VAT rate difference between two bids on the same scope of work is one of the most common discrepancies in a comparison, and one of the easiest to fix once identified.
The reduced rate never depends on a contractor's RGE certification
The 10% VAT rate, applicable to improvement, transformation, fitting-out or maintenance work on a dwelling completed over two years ago, never depends on the RGE (Reconnu Garant de l'Environnement) certification of the contractor submitting the bid. Since 1 March 2025, the only administrative condition for applying that rate is a statement on the quote or invoice by which the client certifies the work meets the required conditions, as detailed on the impots.gouv.fr page on VAT rates for housing works. RGE status governs access to certain energy-renovation grants, never the VAT rate itself.
Why two contractors sometimes apply two different rates to the same site
A contractor applying 20% where the project meets the conditions for the 10% rate is not necessarily charging more overall — they are simply invoicing the wrong rate, often out of excessive caution or unfamiliarity with the actual condition for applying it. Comparing bids without reconciling this point means comparing a correct pre-tax amount with another correct pre-tax amount subject to a different rate, which distorts the final ranking even though the scope of work has not actually changed.
Comparing bids with ChatGPT, step by step
Step 1: Preserve the original bids and create a common item list. This list must exist before any calculation, so that every compared row genuinely matches the same service across every contractor consulted.
Step 2: Separate base scope, alternatives, discounts and taxes into distinct columns. A combined total that blends these categories rules out any reliable comparison, even when the totals look close between bids.
Step 3: Check the VAT rate each contractor applied on items of the same nature. A rate difference on an identical scope must be flagged as a discrepancy to clarify, never silently recalculated by ChatGPT on the contractor's behalf.
Step 4: Identify every scope difference and prepare a written question instead of forcing a ranking. If a bid does not state whether installation is included, the right response is a question to the contractor, never an assumption adopted to complete the table.
Step 5: Apply an official ranking only if scoring criteria have been explicitly supplied. Absent weighted criteria communicated by the client, a summary table must remain an informative comparison, never a ranking presented as a final decision.
Fictional example for testing the method: A quotes €12,000 including installation; B quotes €10,500 without specifying it. The correct request to ChatGPT is to build the comparison and a clarification question on that specific item. Since no scoring criteria are supplied in this example, no official ranking should be produced.
When this comparison is not enough
A well-built comparison table does not resolve every situation encountered when analyzing bids.
A public contract whose amount exceeds the formalized-procedure thresholds requires a documented scoring method communicated to bidders before bids are received, a requirement beyond what a comparison table can establish after the fact: criteria must be set in advance, never reconstructed after seeing the proposals. A very large price gap between two bids on an identical scope — well beyond what the market's usual dispersion explains — should prompt a review of the cheapest contractor's understanding of the CCTP before it is awarded the contract, rather than dismissing or accepting it without clarification. Finally, a comparison prepared with AI assistance remains a decision-support tool: the final decision to award a contractor always belongs to the client, advised by the architect, never to the table itself.
Training to make bid analysis more reliable
A well-structured comparison is not enough on its own: it also needs adapting to every tender, with bids that never quite resemble each other from one project to the next. For an architect practising as a self-employed professional (NAF code 7111Z in France), the FIFPL fund covers part of the cost of Qualiopi-certified training under 2026 criteria set at €300 per day and €900 per year, with e-learning capped at 50% of the daily rate — funding that must be arranged before training starts, not after.
Frequently asked questions
Does a contractor need RGE certification to apply the 10% VAT rate?
No: the 10% VAT rate depends on the nature of the work and the dwelling's age, never on the RGE certification of the contractor carrying it out. Two bids applying different rates to the same scope therefore do not reflect a quality difference, but an application error to clarify before ranking. That clarification should be requested in writing from the contractor concerned, rather than corrected unilaterally in the comparison without their agreement.
Can ChatGPT produce an official ranking of the contractors consulted?
Only if weighted scoring criteria have been explicitly communicated by the client. Absent that, ChatGPT must limit itself to an informative comparison of amounts, scopes and identified discrepancies, without presenting an order of preference as a settled decision. If criteria are supplied partway through the analysis rather than at the outset, the comparison already produced should be redone in full rather than simply reordered.
Does the principle of equal treatment apply to a private client building their own home?
No: this principle, set out in Article L3 of the French public procurement code, concerns public buyers exclusively. A private client retains much wider contractual freedom to choose their own criteria, provided they stay consistent across bids so the comparison remains defensible. That consistency must be applied from the first bid received, since changing criteria partway through the tender weakens the comparison in the eyes of both the client and the contractors consulted.
What should be done if the scope of two bids is not comparable?
Good practice is to prepare a precise written question to the contractor whose scope remains ambiguous, rather than assuming an answer to artificially complete the table. An unclarified item should stay flagged as uncertain until the contractor responds. The final ranking, if one is produced, must then explicitly mention that uncertainty rather than letting it disappear when the result is presented to the client.
A reliable bid analysis rests on three elements that do not change from one tender to the next: a common item list applied without rewriting the original bids, an explicit check of VAT rates applied on identical scopes, and an official ranking reserved for cases where the client has communicated weighted criteria.
Before analyzing bids, our ChatGPT method for the works CCAP details how to set the administrative clauses framing this tender, and our ChatGPT method for DPGF breakdown by trade revisits the price breakdown every bid must answer.
Training 100% fundable via OPCO/FIFPL. Qualiopi-certified programme. To structure your bid analyses with AI as part of our AI for architectural management training, contact Educasium and specify your status (employee, self-employed, business owner) and your goal. This pack is also available on its dedicated ChatGPT skill page.