AI and public tenders: the honest SME guide (what works, what does not)

Until 2023, software tools for tenders were essentially monitoring systems (notice aggregators) and document libraries. The arrival of large language models has changed the game: for the first time, it becomes technically possible for a machine to understand a DCE, extract its key information and generate professional-quality text adapted to the context.

Three technical conditions converged simultaneously. Context capacity: modern models process 100,000-200,000 tokens in a single request — the equivalent of a complete 300-page DCE. French language quality: 2024-2025 models write fluent professional French. Accessible cost: analysing a complete DCE cost €50-100 in 2023, costs €2-5 in 2026.

Not all tasks are equal when it comes to AI. Here is where artificial intelligence brings a decisive advantage.

AI also has its limits, sometimes unexpected. Here are the areas where caution is essential and strong human involvement must be maintained.

Trap 1: using ChatGPT as the main writing tool. ChatGPT is an excellent general-purpose tool, but is not specialised in French public tenders. It produces generic bids without references to the Code de la commande publique, without French administrative terminology.

Trap 2: blindly trusting AI extractions. An AI can make a mistake on a criterion, forget a document or misinterpret a clause. Systematically checking critical extractions against the source document takes 10 minutes and avoids disasters.

Trap 3: submitting a 100%-AI bid without human review. Public buyers are increasingly detecting AI-generated bids. A raw AI bid is weaker than an average human bid. The AI produces a first draft that must be enriched with specific details: team names, precise project references, your own figures.

Trap 4: not updating the company profile. The quality of AI outputs depends directly on the quality of inputs.

Trap 5: running AI without metrics. If you use AI but do not measure its impact, you do not know what works.

Generalist tools (ChatGPT, Claude.ai, Gemini). Useful for occasional tasks: brainstorming, reformulation, proofreading. Price €20-30/month. Limitation: no specialisation in public tenders, no monitoring. Recommended for: SMEs responding to fewer than 3 tenders/year.

Tools specialised in French public tenders (Maître AO and a few emerging competitors). Integrated monitoring, full DCE analysis, technical bid respecting the buyer's imposed framework. Price €39-199/month. Recommended for: SMEs responding to 3+ tenders/year.

Anglo-Saxon tools (Loopio, Responsive, RocketDocs). Excellent for international private B2B (English RFPs), but unsuited to French public tenders. Recommended for: large B2B companies with international activity.

Level 1 — Beginner: automated monitoring + DCE analysis. This is the profitable minimum. You keep human writing but identify opportunities better and analyse DCEs faster. Effort: 5 minutes of configuration. Gain: 5-10 hours/week.

Level 2 — Intermediate: + bid section generation. You use AI to produce the 60-70% of structured sections and humanly enrich the differentiating sections. Gain: ×2 to ×3 on the volume of possible responses.

Level 3 — Advanced: + administrative pre-filling + DECP market intelligence. The entire administrative and analytical workflow is automated. Your team focuses exclusively on strategic judgement and differentiating writing. Gain: ×3 to ×5 on volume.

To materialize the productivity leap AI brings to public tenders, here is the comparison of two construction SMEs of equivalent size (15-20 employees), observed over 6 months in 2026:

The numbers speak for themselves: SME B handles 3x more DCEs, submits 3x more responses, with a better win rate (AI enables better upstream selection). Additional revenue funds 1,700+ times the cost of the AI subscription. This ratio is consistent: across 50+ SME Maître AO users tracked in 2026, minimum ROI is ×30 (starting SMEs) and maximum ×3000 (structured SMEs scaling volume).

Sandra runs a 24-employee private security company in southern France. She has been responding to public tenders for 6 years, mainly hospital and municipal security. Before Maître AO, she had a full-time sales person on bid writing and regularly hired an external consultant at EUR 2,500-4,000 per file on strategic tenders.

The trigger. In January 2026, Sandra discovered Maître AO via a Google search "security technical proposal software". She tested the free analysis on a hospital security DCE she was evaluating. The scoring report identified in 6 minutes what her team had spent 8 hours formalizing, plus DECP data (previous holder, historical EUR 287K/year contract value) shaping her pricing strategy.

Routine usage. Sandra subscribed to the EUR 79/month Pro plan. Over 4 months:

• Response volume: from 3 to 8 per month
• Contracts won: 9 of 32 responses (28%) vs 4 of 12 previously (33%)
• Win rate slightly down (-5 points) but absolute gains doubled
• No more external consultant: net savings of EUR 12,000 over the half-year
• Dedicated sales person redeployed to client follow-up and direct prospecting

Sandra's key insight: "Maître AO does not replace my expertise, it frees up my time. The painful part — extracting CCTP requirements, structuring the proposal, filling DC1/DC2 forms — is processed in minutes. I now spend most of my time personalizing the sections where we genuinely differentiate from competitors."

This profile — SME director in B2B public sector, solid expertise, lean team — represents the majority of Maître AO users in 2026.