Upload your pitch deck, business model, forecast, and data room. A firm of brutal AI analysts gives you investor-style feedback on the narrative, the numbers, the assumptions, the red flags, and the diligence gaps that make VCs pass.
25 analysts · 5 artifacts · brutal, then constructive
Every analyst has one lens and zero patience.
They quote your own words back at you, then hand you the fix.
“Your opener is “the world is changing.” So is the weather.”
“A “$500B TAM” with no source is a number you Googled.”
“You showed me revenue and hid the cost side entirely.”
“Three of these claims won't survive one reference call.”
““AI” is not a moat. Neither is “first mover.””
“This is three companies wearing one trench coat.”
“Your opener is “the world is changing.” So is the weather.”
“A “$500B TAM” with no source is a number you Googled.”
“You showed me revenue and hid the cost side entirely.”
“Three of these claims won't survive one reference call.”
““AI” is not a moat. Neither is “first mover.””
“This is three companies wearing one trench coat.”
“I'm not writing this check until slide 13 has a number.”
“A flat line that snaps vertical at month 9 is a wish.”
“Drop conversion 30% and you're out of runway by Q3.”
“A $99 tier can't average a $14K ACV. Do the math.”
“Your “moat” is a feature with a one-week switching cost.”
“You're pitching Step 3 with a Step 1 that's six weeks old.”
“I'm not writing this check until slide 13 has a number.”
“A flat line that snaps vertical at month 9 is a wish.”
“Drop conversion 30% and you're out of runway by Q3.”
“A $99 tier can't average a $14K ACV. Do the math.”
“Your “moat” is a feature with a one-week switching cost.”
“You're pitching Step 3 with a Step 1 that's six weeks old.”
No vague “tighten the narrative.” Every critique comes with the exact slide, the exact number, and the exact change.
“You’re asking for $20M and the deck doesn’t have a single line of cost structure. Slide 13 shows the ask and nothing else.”
Add a P&L slide with trailing-12 revenue, COGS, opex, and net burn. Compute the burn multiple out loud. Re-state the ask with pre-money, post-money, and dilution.
Each one gets its own specialist firm. Or hand over the whole package and let a partner connect the dots.
Narrative, market, traction, economics, and moat, reviewed slide by slide.
Eight quants stress-test the curve, the funnel, the cohorts, and what breaks if a number misses by 30%.
Pricing, take-rate, channels, and lock-in, pressure-tested on whether the money-making mechanism actually pencils.
Drop one document and the firm judges it for its own job, not for failing to be a full deck.
Hand over everything. Each file goes to its firm, then a partner catches every contradiction between them.
Your fundraising materials are confidential.
Roast My Startup is designed around that assumption.
Every roast becomes one structured fundraising-readiness report you can act on, line by line.
A generic chat box can give you an opinion. Roast My Startup runs the review an investment committee actually would.
The questions founders ask right before they get roasted.
A blank chat box gives you a quick opinion on a single deck. You can absolutely ask ChatGPT or Claude to “act like a VC,” and it’ll produce a useful first pass.
But real investor feedback never comes from one voice. One partner cares about market size, another picks apart GTM, another stress-tests the financial assumptions, another hunts for diligence gaps, another questions the competitive story, and someone always checks whether the deck and the model actually agree.
Roast My Startup is built to mimic that entire committee. It reviews your deck, model, forecast, and data room through 25 investor-style lenses: narrative, market, traction, GTM, financial assumptions, competition, diligence gaps, risk, metric consistency, and the objections a partner will actually raise. It then turns that into a structured pass-risk report.
The workflow was shaped by real VC advisor feedback we got while sharpening our own raise, the same process that turned a decent fundraising package into one advisors immediately recognized as much stronger. The point is speed and depth: pressure-test the whole investor package before investors do.
Anything in the raise. Drop a pitch deck (.pptx, .html, .md), a model or forecast (.csv, .tsv, .xlsx), a business model, or a full data room. Each gets its own specialist firm. The real payoff is the Full Raise, which reviews them together and catches the contradictions between your story and your numbers.
Specific enough to act on. No vague “tighten the narrative.” Every critique quotes the exact slide or number that fails and hands you the fix. For example:
Brutal first, constructive always. You leave with a punch list, not bruises.
Not with certainty. Fundraising hinges on timing, investor fit, traction, founder credibility, and category, which no tool can guarantee. What it does well is catch the common pass-risk issues: an unclear narrative, unsupported assumptions, inconsistent metrics, missing diligence, weak market framing, or numbers that don’t line up. It’s a pressure-test, not a funding oracle. For very niche markets, treat the feedback as a structured outside review rather than the final word on your space.
No. It helps you show up better prepared before those conversations. A human who knows your company and market deeply is still invaluable. This is a fast first-pass investor-readiness review that clears the obvious issues so your time with real advisors and investors goes further.
It’s most useful for pre-seed through Series A founders preparing for angels, accelerators, seed funds, and early-stage VCs. Later-stage teams can still get value from the model and data-room review, but the product is designed around early fundraising readiness.
Trust is the point. We don’t sell your documents, share them with third parties for marketing, or make them visible to other users. Uploaded materials are used only to generate your analysis.
Decks, models, and forecasts are parsed in your browser, so only the extracted text needed for the review is sent. Every saved session is private to your account, and you can delete any of them anytime from Past sessions.
Not ready to share a full data room yet? Start with a deck-only roast or a redacted version.
No. We don’t train any model on your decks, numbers, or feedback, and we don’t sell or resell training data. Your materials are processed through the AI provider’s API (Anthropic or OpenAI) purely to generate your report, and both providers state that data sent through their APIs isn’t used to train their models by default. At most we look at aggregate, non-identifying usage patterns to improve the product, never your confidential business content.
You can bring your own Anthropic or OpenAI key, or run on hosted credits. Either way, you start in seconds. Bringing your own key runs reviews through your own model provider account at the underlying model cost, which technical founders like for the control and pricing transparency. Stored keys are encrypted, tied to your account, never shown back in the browser, and never used for anyone else’s runs.