The first AI-powered platform to sort 223,000 public procurement opportunities

French public procurement is a vast market. In 2024, 223,383 public contracts were recorded with a total value of €233.3 billion—nearly 8% of French GDP—according to the OECP census published in March 2026. It is a commercial opportunity that every serious B2B organisation should explore.

It is also an administrative maze that lawmakers themselves acknowledge is excessively complex. That complexity is precisely what creates the opportunity for us.

The complexity is not imagined; it is an established legal reality. France’s Public Procurement Code, which consolidated half a century of disparate legislation in 2019—the 1975 law on subcontracting, successive European directives, the 2015 and 2018 ordinances, simplification laws from 2020 and 2021, the 2023 green-industry decrees and the simplification measures adopted at the end of 2025—now contains several thousand articles.

In its official guidance, Bercy’s Directorate of Legal Affairs (DAJ) itself acknowledges that the law “appears complex and liable to discourage SMEs and TPEs” and that it “also penalises buyers with regard to legal certainty when awarding contracts” (DAJ, updated in 2024).

Public institutions see the problem. A succession of written parliamentary questions in the Senate has asked the Minister for the Economy about the barriers TPEs and SMEs face when accessing public procurement. The finding is consistent: “TPEs and SMEs do not have the human resources to identify and respond to the many calls for tenders published in their sector” (ministry response, JO Sénat, 8 December 2022).

Despite a series of simplification laws and decrees, outcomes remain uneven. TPEs and SMEs account for 99% of the French economy, yet secure only 60% of public contracts by volume and 30% by value (OECP data cited by the Senate, 2022). Large companies capture 43.7% of total contract value despite representing only 18.6% of successful bidders (OECP, 2023 data).

The latest simplification measures, adopted at the end of 2025, do not address the underlying issue.

Decrees no. 2025-1383 and no. 2025-1386 of 29 December 2025, which entered into force on 1 January 2026, reduce the maximum turnover requirement from twice to one and a half times the contract value and raise the thresholds below which no formal procedure is required: €60,000 excluding VAT for supplies and services from 1 April 2026, and €100,000 excluding VAT for works from 1 January 2026.

These are useful technical adjustments, but they do not change the central operational problem: the candidate company remains solely responsible for the preliminary work of identifying, qualifying and analysing opportunities.

The gap between promised simplification and persistent operational complexity reveals five structural limitations that continue to make tender monitoring expensive and labour-intensive in 2026.

• First, CPV triage remains manual.
  

  Two to three hundred new public consultations are published every day and must be classified across several thousand CPV codes. In theory, an algorithm should be enough. In practice, market tools deliver lists that users must read one by one to decide whether an opportunity matches their business. A significant share of a business development manager’s time is therefore spent on a task the machine should perform unaided.

• Second, DCE files are never retrieved automatically.
  

  A published consultation is merely an announcement; the consultation file—RC, CCTP, CCAP, BPU, AE, plans and appendices—is hosted elsewhere, across dozens of different buyer-profile platforms, each with its own interface. No standard monitoring tool retrieves them automatically at this scale. Companies pay a subscription to receive listings, then spend several hours each week downloading the documents those listings announce.

• Third, analysis occupies one or two full-time equivalents (ETPs).
  

  Once the DCE has been retrieved, reading the RC, CCTP and CCAP, cross-checking the documents, identifying disqualifying clauses, estimating the workload, calculating a defensible price and establishing a credible margin takes two to six hours for each serious opportunity. Across the daily flow, this represents four to eight thousand euros in monthly payroll, or fifty to one hundred thousand euros a year, devoted to tasks the machine should absorb.

  It is an organisational legacy maintained by default because no available tool can yet handle the work.

• Fourth, profitability is never calculated early enough.
  

  The most expensive commercial mistake for an AO bid team is not missing a contract it should have pursued; it is winning a contract it should have declined. A loss-making award consumes resources for months and reduces the capacity to bid elsewhere. No standard monitoring tool estimates profitability in advance at this level of integration. The decision to bid rests on qualitative criteria without a quantified projection.

• Fifth, public institutions acknowledge the impasse.
  

  In a report published in 2023, the Inspection générale des finances estimated that local and regional authorities could save up to ten per cent on their purchases—approximately five billion euros a year—by further professionalising and rationalising their procedures. Administrative complexity creates costs on both sides of the contract: buyers lose negotiating margin, while candidates absorb the cost of preliminary work through payroll.

  It is a deadweight loss for the entire ecosystem.

• Ingestion.
  

  The pipeline ingests publications from France’s Bulletin officiel des annonces de marchés publics (BOAMP) and Europe’s Tenders Electronic Daily (TED). Together, the two sources cover the great majority of French and European public consultations. Other channels, such as notices in the local press, remain at the margins, but their share is residual.

• Smart CPV triage.
  

  The Common Procurement Vocabulary contains more than 9,000 codes organised across five levels of granularity, classifying contracts by type of service or supply. To configure the tool, users describe their business, services, exclusions and sectors to the AI in French. It selects the relevant CPV codes and calibrates the filter.

• Automatic DCE retrieval.
  

  This is where the most visible technical achievement lies. Every public buyer has its own buyer profile and is under no legal obligation to provide the DCE through BOAMP or TED. The operational reality is that more than five hundred buyer profiles must be covered, each requiring dedicated code adapted to a particular buyer-profile platform or family of platforms. The complexity does not stop there.

  Consultation documents must be available to economic operators free of charge, in full, directly and without restriction. In theory, the documentary content of a consultation is therefore open; in practice, it remains technically fragmented. Some buyers have even added CAPTCHAs or other security measures to data that is, by design, public—administrative logic at its finest. This made our task more complex.

  Across all consultations published within our monitoring scope in April 2026, Logiks AO automatically retrieved the DCE in 97% of accessible cases. Every DCE file—RC, CCTP, CCAP, BPU, AE, plans and appendices—is downloaded, classified, indexed and prepared for analysis.

• AI analysis of the consultation rules.
  

  The RC is read first because it contains the disqualifying criteria. The AI extracts the required references, minimum turnover, requested guarantees, mandatory certifications, prescribed team composition, submission deadlines and scoring weights. If any of these requirements does not match the company’s configuration, the file is filtered out before any detailed analysis. There is no value in spending time on a contract the company cannot win.

• AI analysis of the DCE and confidence scoring.
  

  If the RC passes the filter, the AI analyses the complete DCE and extracts the commercial substance: functional scope, technical constraints, expected deliverables, estimated workload and identifiable risks. This is where one of the tool’s most distinctive features comes into play. Before producing an estimate, the pipeline calculates a documentary confidence score: does the DCE contain enough information to support reliable pricing, or are critical elements missing?

  If the confidence score is too low, the file is marked as requiring clarification from the contracting authority before any costing begins. That discipline is rare. Most AO bid teams price incomplete DCE files blindly and discover the gaps during delivery. The tool refuses to price a file it does not understand well enough.

  If the documentation is incomplete, it advises whether questions should be submitted to the buyer.

• Proposal and budget scoring.
  

  For files that pass the confidence filter, the AI produces a structured cost estimate: breakdown by lot, person-days by role, proposed selling price and expected gross margin. The estimate is not a client deliverable; it is an internal decision-making tool. It informs the “bid or no bid” decision and provides a starting point for the actual commercial proposal if the decision is positive.

• Profitability-based prioritisation.
  

  Using the cost estimate and expected margin, the file receives a final priority score. The most profitable opportunities rise to the top of the table; those with a low or negative expected margin move down. Opportunities are no longer ranked by publication date, contract size or the business development manager’s intuition, but by estimated profitability.

  No standard monitoring tool reverses the process in this way at the same level of integration, and it is probably Logiks AO’s most important economic lever.

The daily triage interface

Above this machinery sits a deliberately simple interface designed around the daily routine of a company director or business development manager. A morning dashboard presents new opportunities ranked by priority score. Each opportunity appears on a card with the essential information: title, contracting authority, deadline, confidence score, estimated margin and a short AI summary.

• One gesture, one decision.
  

  With a swipe, the user can move the opportunity to bid preparation, request additional information or archive it. It is a Tinder-style interface applied to commercial triage: minimal interaction, a binary or three-way decision and maximum execution speed. Work that once required two to three hours a day of scanning poorly digitised PDFs can now be completed in a five- to ten-minute morning session.

  The serious intellectual work—drafting the response, calibrating the commercial position and conducting internal negotiations—begins with files that have already been filtered, costed and prioritised.

Calculating the return on investment for Logiks AO does not require a complex study. The value is visible directly in the payroll of organisations that actively bid for public contracts.

Across most consulting firms, IT services firms (ESNs), engineering consultancies, specialist agencies and large-company AO bid teams, one or two people now work full-time on the preliminary tasks the tool automates: CPV triage, DCE downloads, RC review, extraction of technical constraints, workload estimates and profitability calculations.

At the average salary of a business development manager in France, this represents four to eight thousand euros in monthly payroll, or fifty to one hundred thousand euros a year. Logiks AO absorbs most of that workload.

The benefit extends beyond payroll savings. Business development managers released from mechanical tasks can focus on the work that creates genuine value: drafting the offer, calibrating the technical proposal, conducting internal negotiations and maintaining relationships with recurring contracting authorities. The role moves up the value chain. Teams stop chasing the flow of notices and regain control of their bidding strategy.

There are three direct effects.

• First, a higher file-qualification rate.
  

  The opportunities that proceed to a bid are those that have successively passed the CPV filter, the RC filter, the documentary-confidence filter and the profitability filter. Files are rejected at every stage. Those that remain have already been assessed by the machine, giving the human commercial team a solid basis from which to begin.

• Second, a stronger average margin on winning bids.
  

  Profitability-based scoring systematically directs bidding effort towards the most profitable opportunities. Teams stop winning contracts that consume resources without providing adequate financial return. They may win fewer contracts, but they win the right ones.

• Finally, a shorter cycle time.
  

  Five minutes of morning triage replaces two to three hours of daily scanning. The AO bid team recovers time for high-value work and can respond to a larger number of consultations without hiring additional staff.

Logiks AO is an internal tool designed for our own use. It now represents a competitive advantage that we have chosen not to dilute, so we make it available only to trusted partners. If the tool is relevant to your organisation, we can test it on three to five real DCE files alongside your current method and compare the time spent, risks detected, files rejected and margins estimated.