A typical Belgian SME spends 15 to 40 hours on a serious proposal. For public tenders, that number can go higher — 60 or even 80 hours when you count the back-and-forth between technical teams, legal review, and formatting.
Most of those proposals lose. Win rates of 10–20% are common. Which means for every contract you win, you’ve invested hundreds of hours in bids that went nowhere.
AI doesn’t change the win rate. A badly positioned offer with the wrong pricing will lose regardless of how slickly it’s written. But AI changes the economics: it can cut the time to produce a proposal by 50–70%, which means you can bid on more opportunities or spend more time on strategy instead of wordsmithing.
Here’s what’s actually changing — and what isn’t.
What AI does well in proposal writing
First drafts in minutes instead of hours
The blank page is the enemy. Most proposal writers spend the first two hours just getting a structure in place and producing a rough version that they can then refine.
With Claude or ChatGPT, you can paste the tender requirements and get a structured first draft within minutes. Not a finished product — but a solid foundation that captures the right sections, addresses the evaluation criteria, and uses appropriate language.
The prompt that works:
Here are the requirements for a public tender [paste the specification].
Write a first draft of our technical proposal. We are a [describe your company]
with experience in [relevant experience]. Address each evaluation criterion
explicitly. Flag any areas where you're unsure about our capabilities.
That last sentence is crucial. You want the AI to be honest about gaps rather than inventing capabilities you don’t have.
Tailoring boilerplate to specific contexts
Every company has standard paragraphs — about their methodology, quality management, team composition, references. The problem is that pasting the same boilerplate into every proposal is obvious and uninspiring.
AI excels at taking your standard text and adapting it to a specific context. Feed it your boilerplate about project management methodology, tell it the tender is for a hospital IT system, and it’ll rewrite the paragraphs with healthcare-specific language and relevant examples.
This is genuinely useful because it solves a real problem: the tradeoff between efficiency (reuse existing text) and quality (tailor everything). AI lets you have both.
Compliance checking
Public tenders are notoriously rigid about formal requirements. Miss a required annex, use the wrong terminology, or forget to address a specific evaluation criterion, and you’re out — regardless of how good your offer is.
AI is excellent at cross-checking. Upload the tender specification and your draft proposal, and ask:
Compare my proposal against the tender requirements. List every requirement
from the specification and indicate whether my proposal explicitly addresses it.
Flag anything that's missing or only partially covered.
This catches things that human reviewers miss under deadline pressure. It’s not infallible — AI can misinterpret ambiguous requirements — but it’s a thorough first pass.
Summarising tender documents
Before you can write a proposal, you need to understand what’s being asked. For complex public tenders, the specification documents can run to hundreds of pages, with technical requirements buried in annexes and evaluation criteria scattered across multiple documents.
AI can condense this into a structured brief: what’s being asked, what are the evaluation criteria and their weights, what are the mandatory requirements, and what are the key deadlines. This gives your team a clear starting point in minutes instead of the hours it takes to manually digest a thick specification.
What AI does badly
Strategy and positioning
AI cannot decide whether you should bid on a tender. It doesn’t know your real capacity, your actual relationship with the client, your competitor’s likely offer, or your strategic priorities for the quarter.
The go/no-go decision is still entirely human — and it’s arguably the most important decision in the entire proposal process. Bidding on the wrong opportunities is far more costly than writing a slow proposal for the right one.
Pricing
AI can generate a pricing structure or a cost breakdown template, but it cannot tell you what to charge. Pricing depends on your costs, your margins, your competitive position, the client’s budget expectations, and a hundred other factors that AI has no access to.
If you ask AI to “suggest a competitive price”, you’ll get a generic number pulled from nowhere. This is dangerous — pricing is where proposals are won or lost.
Understanding what the client actually wants
Behind every formal specification, there’s a human with priorities, preferences, and concerns that aren’t written down. The tender might ask for “a robust data platform” but what the client actually cares about is avoiding the disaster they had with the previous supplier.
AI reads the text. It doesn’t read the room. Proposals that win usually demonstrate an understanding of the client’s real situation — and that understanding comes from conversations, site visits, and relationships, not from processing documents.
References and case studies
AI can write a case study. But it can’t write your case study — not with the specific numbers, outcomes, and client context that make references credible. The moment a evaluator notices that your reference reads like generic marketing copy, your credibility drops.
Use AI to structure and polish your case studies, but write the substance yourself. The specific details are what make references believable.
The practical workflow
Here’s a proposal workflow that integrates AI at the right points:
Day 1: Analyse and decide (human)
- AI summarises the tender specification into a structured brief
- Team reviews the brief and makes a go/no-go decision
- If go: define the win strategy, pricing approach, and team allocation
Day 2-3: First draft (AI + human)
- AI generates a structured first draft based on the specification and your company profile
- Technical experts review and correct their sections
- AI checks compliance against the tender requirements
Day 4-5: Refine and finalise (human + AI)
- Team refines the strategy, pricing, and technical content
- AI polishes language, checks consistency, and formats the document
- Final human review and sign-off
What used to be a 2-3 week process can compress into a single week. Not because the strategic work goes faster, but because the mechanical work — drafting, formatting, compliance checking — shrinks dramatically.
For public tenders specifically
Belgian public procurement has specific features that make AI particularly useful:
Standardised evaluation frameworks. Most public tenders use weighted criteria (e.g., price 40%, technical quality 30%, approach 20%, references 10%). AI can help you allocate word count and emphasis proportionally to the weights — something many bidders get wrong.
Formal language requirements. Public tenders in Belgium often require proposals in Dutch or French. AI handles both languages well and can translate or adapt proposals between languages, which is especially useful for federal tenders that accept multiple languages.
Large specification documents. It’s not uncommon for public tender specifications to exceed 100 pages. AI’s ability to summarise, extract key requirements, and cross-reference compliance is particularly valuable here.
Volume. Companies that actively bid on public tenders often track dozens of opportunities and submit multiple proposals per month. The time savings from AI compound quickly at this volume.
If you’re tracking public tenders systematically, tools like TenderWolf can help you find relevant opportunities before the AI-assisted proposal process even begins.
The honest assessment
AI won’t turn a mediocre bidder into a winner. The companies that win proposals do so because of strong positioning, competitive pricing, credible references, and a genuine understanding of the client’s needs. None of that changes with AI.
What AI does is remove the excuse of “we didn’t have time to bid on that.” When a proposal takes 15 hours instead of 60, you can pursue opportunities that you would have previously ignored. And you can spend more of your time on the parts that actually influence the outcome — strategy, pricing, client relationships — instead of the mechanical production of documents.
That shift, from proposal writing as a bottleneck to proposal writing as a routine process, is the real change.
What’s next?
- Want to try writing a proposal with AI? Getting started with Cowork shows you how to set up Claude for document creation
- Looking for public tenders to bid on? TenderWolf monitors European public tenders with AI matching
- Interested in the broader AI tool landscape? Read our AI tools overview for Belgian SMEs