No spin · the cynical questions, answered

You’ve already thought of the hard questions. Good — here are the honest answers.

These are the real objections sophisticated buyers raise about WinGrants AI — pulled from actual sales calls. We’d rather answer them in the open than dodge them on a call, and where the honest answer is “no,” we say no — and explain why that’s fine.

Show the work Every answer links to the proof behind it — a sample, a benchmark, a data-flow diagram — not a reassuring adjective.
No fake victory laps We won’t claim named wins we haven’t earned or training data we don’t have. Early-stage, stated plainly.
Filter to your fear Pick the worry that’s actually blocking you. The privacy slider lets you see exactly what leaves your perimeter.
The objections

Pick the question that’s actually stopping you.

Privacy & IP

“We’d need assurance the data we add does NOT leave the company. That’s a must — it’s the only red flag.”

What you’re really worried about

Ten partners’ confidential strategy leaking — or quietly training someone’s model.

The honest answer

Fair — this is the one objection that should block a deal until it’s answered. We separate three things most tools blur: where the application runs, where your data is stored, and where model inference happens. You choose the inference route. On Enterprise Campus Compute, inference runs on your own machines and nothing leaves your perimeter at runtime — air-gap capable. ISO/IEC 27001:2022, isolated database, bring-your-own-key. The slider below shows exactly what leaves at each mode.

Data-flow diagram · Trust & Security hub

Raised by Applus+ / IDIADA

Privacy & IP

“Is this GDPR-compliant and suitable for an EU public research institution?”

What you’re really worried about

Procurement and legal need contract-grade clarity, not a badge.

The honest answer

Compliance is a data flow and a contract, not a logo. We publish data residency, sub-processors, retention, deletion, access controls and a Data Processing Agreement, and offer on-premise or restricted deployments where a managed route won’t pass review. WinGrants AI is operated by Healthdev OÜ — a registered Estonian (EU) entity.

GDPR FAQ · DPA template · sub-processor list

Raised by universities & RTOs

Why not ChatGPT

“The only difference between you and ChatGPT is that you also look for partners, right?”

What you’re really worried about

Why pay you for something my €20 tool already does.

The honest answer

No. The moat isn’t the model — it’s the Horizon workflow around it: call decomposition into requirements and implied priorities, structured section drafting, a multi-model evaluator mapped to official criteria, redraft loops that fix substance not prose, and CORDIS consortium logic. You can’t prompt that in one window — and we maintain it as the templates and rules change.

Side-by-side: generic model vs WinGrants, same call

Raised by an Icelandic university

Why not ChatGPT

“Did you train on private winning proposals? If not, how can it know what wins?”

What you’re really worried about

Only funded proposals hold the winning pattern.

The honest answer

We didn’t, and we won’t claim we did. We engineer context from public call texts, official templates and evaluation criteria. Most proposals don’t fail because someone lacked secret examples — they fail obvious criteria, weakly. Catching the weak answers is the job.

Data-sources map · rules traced to official criteria

Raised by research managers

Why not ChatGPT

“What stops OpenAI, Anthropic, or the Funding & Tenders Portal from building this?”

What you’re really worried about

No defensibility — a big lab ships this next quarter.

The honest answer

Vertical AI wins on workflow, not foundation models. The defensibility is grant-specific rubrics, instrument-by-instrument logic, the evaluator-calibration loop, CORDIS integration, and the boring operational adoption inside teams — none of which a general model vendor wants to own.

Moat narrative diagram

Raised by investors & strategic buyers

Price & ROI

“I’d rather pay €20/month for ChatGPT than €1k per submission.”

What you’re really worried about

Anchored to consumer-AI pricing.

The honest answer

Different unit. €1,000 is per submission-ready proposal — compare it to a €10k–100k consultant, or to the cost of one preventable rejection on a €4M call. Write one proposal a year and a chat subscription is cheaper; if a rejection costs you the grant, it isn’t.

ROI calculator: hours saved + consultant cost avoided

Raised by DVP Solar

Price & ROI

“15K/year is a hard sell to a manager — is there a lighter way to start?”

What you’re really worried about

Seasonal cashflow and internal budget approval.

The honest answer

Reasonable — so Self-Service stays pay-per-proposal with no commitment. A team can prove value on one real call before anyone signs an annual line item. We’d rather your champion walk into the budget meeting with a benchmark than a brochure.

Single-call pilot + ROI one-pager for the champion

Raised by Applus+

Price & ROI

“Writing is ~25% of the effort — coordinating 60+ partners is the real challenge.”

What you’re really worried about

Drafting speed isn’t their actual pain.

The honest answer

Agreed — and we don’t pretend faster drafting is the whole story. We target the consolidation: consortium mapping, role and budget architecture, contradiction detection across partner inputs, and change-aware redrafting. The bottleneck is messy multi-partner integration — that’s the part we de-risk.

Messy-input → structured-output demo

Raised by Robert Eller

Can we trust you

“You’ve no experience — you’ve never drafted a proposal?”

What you’re really worried about

An outsider can’t build a credible grant tool.

The honest answer

We don’t lead with the founder’s CV, because it shouldn’t carry the sale. We bought the context the hard way — 60+ real trials — and grounded the system in EU evaluator guidelines, not vibes. Judge the output on a hard call of your choosing, not the bio.

Founder note + side-by-side on your own call

Raised by the University of Bologna

Can we trust you

“How much track record do you have? Your company is new.”

What you’re really worried about

New-vendor risk — is this even a real business?

The honest answer

We launched recently and we won’t fake a victory lap — no invented logos, no named full-cycle wins we haven’t earned. What we can show: ISO/IEC 27001:2022 certification, a registered EU entity (Healthdev OÜ), demos run, and measurable before/after deltas from pilots. Trust should be earned by showing the work.

Pilot evidence pack + ISO certificate

Raised across multiple early calls

Can we trust you

“New EU rules ban third-party AI tools to draft proposals — you didn’t know?”

What you’re really worried about

Using us might be non-compliant or reputationally risky.

The honest answer

The rules people cite restrict EU evaluators using AI in peer review — not applicants using AI to prepare a submission. Applicants stay responsible for accuracy, originality and ethics either way. We track the guidance, and for institutions that want hard control we offer restricted and on-premise modes.

AI-use policy explainer + the actual guideline text

Raised by an Icelandic university

Is the AI any good

“This is AI judging itself, not a human evaluator. How did you calibrate?”

What you’re really worried about

The evaluation is circular and not credible.

The honest answer

The score is a diagnostic, not a verdict — its job is to expose weaknesses before you submit, not to predict the panel. It’s calibrated against official evaluator criteria, run by a multi-model committee to surface disagreement, and benchmarked on overlap with real Evaluation Summary Reports. Trust the red flags more than the number.

ESR-overlap benchmark + scorecard with caveats

Raised by Stefano Utili

Is the AI any good

“How do you stop the tool creating a proposal that looks polished but is strategically weak?”

What you’re really worried about

Polish hides weak consortium logic, poor fit, shallow ambition.

The honest answer

We separate writing quality from funding strategy and check the strategy first: call fit, consortium fit, state-of-the-art gap, work-plan credibility, impact pathway and likely evaluator objections — before any redrafting. The most dangerous draft is the one that reads beautifully and says nothing.

Polished-but-weak vs strategically-strong example

Raised by coordinators & senior writers

Is the AI any good

“Who is responsible if the tool inserts a false claim, wrong citation, or non-compliant statement?”

What you’re really worried about

Liability quietly transferred to them.

The honest answer

You’re responsible for what you submit — so we build for that, not against it. Research notes are source-grounded, citations are flagged for verification, and we don’t position output as submission-ready without human review. AI should cut review burden, not remove accountability.

Citation-verification workflow + QA checklist

Raised by consultancies

Will it replace us

“Will AI dilute my scientific voice or make my proposal sound generic?”

What you’re really worried about

Losing authenticity and precision.

The honest answer

You own the science. The tool drafts structure and clarity; section-level editing and human review keep the claims yours. The test is simple — you should recognise your own science in the final draft. Generic AI text is exactly what evaluators punish, so we push for specificity, not polish.

Voice-preservation before/after example

Raised by principal investigators

Will it replace us

“Who actually writes the grant? You don’t contact partners — it’s still me?”

What you’re really worried about

Expected a done-for-you agency.

The honest answer

It’s software, not an agency, and we’re explicit about the split: we take you from 0 to roughly 80–90% — research, first draft, evaluation, redraft — and you own the last 10–20%: scientific truth, partner commitments, final submission. We’re also shipping contact-person discovery and outreach drafting to shrink your last mile.

“What we do vs what you do” graphic

Raised by Stefano Utili

The privacy answer, in detail

Your data path — drag to see what leaves your perimeter.

Best Quality Mode

Commercial frontier models

Where inference runs
What leaves your perimeter
Best for
Available on

Still skeptical? Good.

The fastest way to settle any of these is one hard call and one honest comparison. Run a 90-minute pilot on a real call — one concept, one research note, one section draft, one evaluation, one redraft — and judge the delta yourself.

Book a 90-minute pilot Start free, no card