PRACTICAL AI FOR RFP WRITING
Stop Losing Bids to Paperwork, Not Price.
Most bids aren't lost on price or capability. They're lost on a missed mandatory requirement, a technical proposal that doesn't quite answer the evaluation criteria, or a pricing sheet that arrived in the wrong format three minutes before the deadline.
We set up a full bid workflow inside your team, from the first read of the tender documents to a submission-ready answer, with AI doing the heavy reading, compliance and checking work while your team makes the calls that decide who wins: strategy, positioning and price.
Free open-source Bid Desk or a 1 to 2 week remote setup · Follows APMP best practice · In English · Dutch tenders welcome
Enhances your team
Your bid writers stay in charge of strategy, price and the final text. AI takes over the reading, structuring and checking.
You keep the setup
Bid Desk, your content library and prompts live in your own Claude account and stay with your team after we leave.
Your choice of support
Take Bid Desk free from GitHub and set it up yourselves, or have us set it up with your team. No retainer required.
Who Is This For?
Built for bid teams: bid writers, tender managers, tender specialists and proposal writers at companies that bid on public tenders (Dutch aanbestedingen and EU tenders) or large commercial RFPs: IT and software, engineering and infrastructure, facility and staffing services, healthcare, energy, and consultancies. Also for consultants and agency owners who write their own bids. It suits small and mid-size bid teams best, who want a strong process without a full tender platform subscription.
If a document lands on your desk with a submission deadline, a page limit, a mandatory-requirements checklist, and an evaluation matrix attached, and you've built more than one proposal by copying last year's and hoping you caught everything that changed, this setup is for your team.
We set up practical workflows for turning your own track record into a reusable capability library, building a compliance matrix automatically instead of by hand, turning specialist input into text that scores, and catching the mistakes an evaluator would catch before you submit.
We work in English. Bring tenders in Dutch or English: the AI reads and drafts in both, and your team keeps control of the final Dutch text.
What We Set Up
Five parts of the bid process, set up on your own tenders and content and run by your team on a live tender.
Stop Chasing Tenders You Can't Win
The Pain: Days lost on a bid before you notice a knock-out requirement on page 140, a reference that is one year too old, or a week spent copying the tender into your go/no-go spreadsheet before anyone can decide.
The Fix: Pull the exclusion grounds, suitability requirements, knock-out criteria, award weighting, contract risks and deadlines out of the tender in minutes. Your own go/no-go checklist is filled in with a source for every line, plus the nuances worth discussing, so the decision with your manager takes an hour, not a week.
Read a 200-Page Tender in an Hour
The Pain: Reading and cross-checking long tender documents by hand, then finding the contradictions after the clarification deadline has passed.
The Fix: One pass that builds your compliance matrix (every requirement, wish, question and form, with its reference and owner), a scoring map of each award criterion with its weight and page limit, a backwards plan from the deadline, and draft clarification questions (Nota van Inlichtingen).
Get the Knowledge Out of Your Specialists' Heads
The Pain: Specialists who are too busy to write, and input that arrives late, unstructured and impossible to score with.
The Fix: Generate an interview guide aimed at what the scoring scale rewards, run a short interview, and turn the transcript into structured content: approach, roles, KPIs, risks with mitigations and evidence, saved to a reusable content library for the next bid.
Write the Quality Plan That Scores
The Pain: Generic plan van aanpak text that could belong to any bidder, over the page limit, with promises an evaluator can't measure.
The Fix: Agree your win themes, then write each quality section in the structure of the question: buyer benefit first, every promise SMART, evidence from your own verified track record only, trimmed to the exact page limit.
Score It Like the Evaluator Before They Do
The Pain: Submitting with a knot in your stomach, and only finding out why you lost at the debrief.
The Fix: Run a strict evaluator simulation that scores your draft against the published award criteria and says exactly what is missing for the next score level. Rewrite, score again, then check that text, planning, CVs and price all agree before you submit.
Built on APMP Best Practice
Our workflow follows the bid lifecycle set out by APMP, the global association for bid and proposal professionals. If your team already works to APMP methods, Bid Desk fits the way you work. If it doesn't yet, your team picks up the same habits that certified bid managers use.
Bid/No-Bid Decision
Every tender checked against your track record, with knock-out requirements, key terms and risks flagged before anyone writes a word.
Proposal Planning
A compliance matrix, a scoring map per award criterion and a schedule worked back from the deadline, all from one pass through the documents.
Win Strategy
Win themes tied to evidence from your own content library, carried into every high-weight section.
Content Development
Specialist interviews turned into structured content, and sections drafted to the structure of each question by your team or by AI from your library.
Red Team Review
Every section scored the way the evaluator will score it, with a list of exactly what is needed to reach the next score.
Submission and Lessons Learned
The buyer's own forms filled and checked, then scores and debrief comments fed back into the library for the next bid.
APMP is a registered trademark of the Association of Proposal Management Professionals. Practical AI is independent of APMP and is not endorsed by it.
Bid Desk: Yours to Keep
Bid Desk is our free, open-source app for Claude, shown below. It tracks every bid on one board, saved for your whole team, and checks each tender for eligibility, key terms and compliance, with every point linked to its page. It fills your own go/no-go checklist, and keeps one sheet of bid facts (price, team, dates, promised KPIs) that flags any section still using an old value. For each section, your bid manager chooses who writes it: a team member, or the AI drafting from your own content library in your own voice. It builds the price from your rate card, fills the buyer's own Word and Excel forms, and turns every debrief into a win/loss view of where you lose points.
3-minute walkthrough of Bid Desk on a real Dutch tender, Gemeente Hardenberg, with subtitles.
See each part up close
Three short walkthroughs, each under a minute.
AI answers in the main walkthrough were generated by Claude from this tender and replayed for the recording. The bidder, its references and CVs are fictional.
See what Bid Desk produces:
Download a sample go/no-go checklist (Excel)Filled by Bid Desk on the public WHO tender RFP 079-2026, for a fictional bidder.
Free and open source
MIT licence on GitHub, free for any organisation to use and adapt.
No subscription
It runs in your own Claude account. Nothing to renew.
Yours to adapt
Your team can change the prompts, criteria and templates.
Try It on a Past Tender
The quickest way to judge Bid Desk is on a tender you know well. Send us one you bid on in the last year, and we'll show you in 30 minutes what it finds. Free, and no commitment.
Send one tender
The tender documents are enough, and a public tender from TenderNed needs no NDA. Add your submitted answer and your scores if you want the evaluator comparison, and we'll sign an NDA first.
We run it through Bid Desk
Go/no-go, key terms, compliance matrix and clarification questions, each with the page it comes from. With your answer and scores, the evaluator simulation too.
A 30-minute call
We walk you through the results on screen. You see where it matches your own reading, where it found something you missed, and how the evaluator score compares with your real one.
We delete the documents after the call unless you ask us to keep them.
Why This Is Different
Writing a bid the traditional way is sequential and manual: draft the narrative, build the price separately, cross-check by hand against a requirements list you made yourself. Bringing deep-context AI into that process shifts the work from manual writing to strategic directing and fast iteration, without lowering the bar an evaluator holds you to.
The Old Way
Skimming a long tender and hoping you caught every requirement.
The Practical AI Way
A full compliance matrix and scoring map from one pass through the documents.
The Old Way
Chasing specialists for input that arrives late and unusable.
The Practical AI Way
Short, targeted interviews turned straight into structured, reusable content.
The Old Way
Everyone working in their own AI chat, then chasing changes by email and Excel.
The Practical AI Way
One shared bid: one set of bid facts, a clear view of who owes what, and every section re-checked when a figure changes.
The Old Way
Generic boilerplate an evaluator has read a hundred times.
The Practical AI Way
Quality text built on your own verified track record, in the structure the evaluator scores.
The Old Way
Text, planning and price that drift apart under deadline pressure.
The Practical AI Way
A consistency check across every document before you submit.
The Old Way
Finding the weak spots at the debrief.
The Practical AI Way
A strict evaluator simulation days before the deadline, not after the award.
Two Ways to Work With Us
Take Bid Desk free and set it up yourselves, or have us set it up with your own tenders and content. Either way it runs in your own Claude account and stays with your team.
Free: set it up yourselves
Take Bid Desk from GitHub and publish it in your own Claude account.
- Bid Desk app: bid board, your own go/no-go checklist, key terms, compliance, bid facts, AI or team drafting, pricing, evaluator scoring and win/loss
- Fills the buyer's own Word and Excel templates
- Setup guide, sample outputs and a 3-minute setup video
- Open source under the MIT licence, free for any organisation
Best for teams comfortable setting up AI tools themselves.
We set it up with you, 1 to 2 weeks
We set Bid Desk up on your own tenders and content and train your team on it. Delivered remotely.
- Week 1: an audit of how your team bids today
- Week 2: your content library and Bid Desk set up, and live training on a current tender
- A check-in call after your first bid
- Optional quarterly tune-up as tender rules and AI tools change
Pilot places at a reduced rate. In-person kick-off available in the Netherlands.
See how a setup runsTell us about your team and your next tender, and we'll recommend where to start.
Setting it up yourself?
Watch how to go from GitHub to your team's first bid.
3-minute setup video, with subtitles.
How a Team Setup Runs
Two weeks and around five to six days of our time, built around a real tender your team is bidding on. The setup runs over video calls and shared files, with an optional in-person kick-off in the Netherlands.
1-minute overview of a bid team setup.
Before we start
Preparation
We sign an NDA and data processing agreement before the setup begins, since you'll be sharing past bids, pricing and CVs.
Week 1
Audit
You share 3 to 5 past tenders with your responses, scores or debriefs where you have them, and your current go/no-go checklist. We map what AI can automate, what it can draft from your past content, and what stays with your writers, and walk you through it on one call.
Week 2
Setup and training
We build your content library and set up Bid Desk in your own Claude account over screen share, then train your team live on a current tender.
A month later
Check-in
A call after your first bid with the new workflow, to fix any gaps.
Fixed at two weeks, ending with handover on the last day.
How We Handle Your Data
Bids hold your prices, margins, CVs and client details. Here is how Bid Desk and our setup keep them under your control.
01
It runs in your own Claude account
Bid Desk is published in your organisation's Claude account. We don't host it, and your bids and content library are visible only to the people you share it with.
02
No training on your documents
We set your team up on a Claude business plan, where your documents and chats are not used to train models.
03
Tender files stay on your side
Bid Desk doesn't store the tender PDFs. Each person attaches them in their own browser to open source pages and ask the tender.
04
Agreements before we see anything
An NDA and a data processing agreement are signed before the audit, since you'll share past bids, pricing and CVs.
05
We keep as little as possible
We work from the documents you choose to share, and delete our copies at the end of the setup.
06
Open code your IT team can check
Bid Desk is open source under the MIT licence, so your IT or security team can review exactly what it does before you use it.
Questions about security? for a copy of our data processing agreement before you decide.
Learn From Every Bid You Win or Lose
In public tenders, every award decision comes with the scores and the reasons behind them. Most teams read them once, file them, and repeat the same weak spots on the next bid. As part of the team setup, we turn them into a routine your team keeps running.
Find the pattern
Scores from your won and lost bids, grouped by theme across bids in Bid Desk: which criteria cost you points, which sections score consistently, and what to change in your content library.
Feed the content library
Sections that scored well become the starting point for the next bid. Weak sections are flagged with what the evaluator was looking for.
Sharpen the evaluator
Bid Desk's evaluator prompts are tuned on your real scores, so the simulation marks your drafts the way buyers in your market do.
Track your win rate
Win rate and average quality score in Bid Desk, updated every time your team records a debrief, so you can see whether the changes are working.
Your team adds each new award decision as it comes in, so the learning carries on after we leave.
FAQ
General Questions
Is Bid Desk really free?
Yes. Bid Desk is open source under the MIT licence, free for any organisation to use and adapt. Download it from GitHub and publish it in your own Claude account. Each person who uses it needs their own Claude plan. If you'd like it set up on your own tenders, with your team trained, that's our paid setup.
Does this follow APMP methods?
Yes. The workflow follows the APMP bid lifecycle, from the bid/no-bid decision through planning, writing, review, submission and lessons learned. It works alongside APMP training and certification, and it gives teams without that training a proven structure to work in.
Is the setup in English or Dutch?
Both. Bid Desk has a working-language switch, so analysis and drafts come in English or Dutch. Quotes stay in the tender's own language, and every quote and page has a "Show in English" or "Translate page" button. Drafts follow the tender's language by default and you can change that per bid.
Can the team setup be done remotely?
Yes. The whole setup runs over video calls and shared files: the audit, the content library, Bid Desk in your own Claude account, and live training on a current tender. If you're in the Netherlands and would like an in-person kick-off, that's available too.
Can the whole team work on one bid?
Yes, and there's no shared Claude account. One person publishes your Bid Desk and shares it from Claude's Share menu, with edit access for the people who work on bids. Everyone then works in their own Claude account on the same board, library and bids. If two people edit the same bid at the same moment, the last save wins, so agree who owns each section. Export a backup from the content library regularly, and before the person who owns it leaves, so you can move it to a new owner.
Will this replace our bid or proposal writers?
No. We set up AI to handle the reading, structuring, chasing and checking work. The decisions on which tenders to pursue, how to position against competitors, the price, and the final text stay with your team.
Do we need to use Claude specifically?
Bid Desk runs in Claude, which our team uses daily. If your organisation only allows Microsoft Copilot, the method itself (go/no-go, compliance matrix, content library, evaluator scoring) still works there without the app, and we can help you make the case for a Claude business plan for the bid team.
Will AI make bid writing easy?
Bids still take real work, and they should. We set up Claude to handle the heavy lifting and catch problems before an evaluator does, giving your team more time for strategy and positioning.
Can AI get things wrong?
Yes, and every draft still needs a human check, especially for numbers, capability claims, and anything that could be read as a compliance commitment. We build a red-teaming step into your workflow to catch mistakes before an evaluator does.
How does this differ from standard online AI tutorials?
Most tutorials cover generic prompts or polishing prose. We set up a full workflow on your own tenders and content: qualifying a tender, building the compliance matrix, turning specialist input into content, writing to the award criteria, and scoring the result like an evaluator. Your team runs it on a live tender before handover.
Which Claude plan does our team need?
A Claude business plan for everyone on the bid team, which does not use your documents for training. We help set it up before the start.
Technical & AI Capability Questions
How do we handle strict page, word, or format limits using Claude?
Claude is strong at multi-constraint formatting. We configure prompts to trim drafts to an exact page or word count without losing the claims and structure an evaluator is scoring against.
Does this cover pricing?
Yes. Bid Desk builds the price per deliverable from your own rate card and cost norms, does the sums, and flags the contract terms that change your price. The price decision stays with your team, and Bid Desk checks that price, staffing and planning match what the quality text promises.
What about confidential and commercially sensitive information?
Bids routinely involve your own cost structure, margins, and client-specific details. We set up business-grade workspaces that do not use your documents for training, and mask commercially sensitive figures before feeding past bids or pricing into the model.
Practical Application Questions
Will a bid written with AI read as generic to an evaluator?
Yes, if you ask a chatbot to write a proposal from a generic prompt. Evaluators who read dozens of bids already spot that instantly. We build a permanent library of your own verified track record so AI assembles drafts from your data. You keep full editorial control.
Can these workflows be adapted to other AI tools like ChatGPT or Gemini?
Yes. We use Claude because it handles long documents well, and the method (content library, compliance matrix, specialist interviews, evaluator scoring) applies directly to other mainstream tools.
How much time does this actually save on a typical bid?
The biggest savings come from reading and analysing the tender, building the compliance matrix, and turning specialist input into usable text. We set up these steps to free time for the strategy and positioning that actually win the bid.