A note before this one. The newsletter has been quiet for a few weeks. The cadence in the future will be slower — every two to three weeks, sometimes less. I'd rather ship well than ship often.
For work, I go through a lot of decks.
Not in a fund-of-funds way. In a "founder sent it Tuesday, partner forwarded it Wednesday, someone's old friend pinged me Friday" way. The flow doesn't stop. Most of them deserve 90 seconds, not 90 minutes.
I come from an analytical background. So my default with any deck is the same. Skim it, run my own checks and balances, make notes, ask questions. Useful. Slow.
Early last year, I decided to build an engine to do the first pass for me.
It's the third project I've documented in this newsletter. If you're new, the previous two are the personal CFO who audits my credit cards and the nutrition agent who holds me more accountable than any app.
This one runs my deal flow. I call it The Analyst.
What it does
I gave a Custom GPT a system prompt loaded with my own analytical framework. The same checks I'd run manually, just structured. Then I told it to go one step further and pull whatever it could find on the public web. Funding history, founder background, ecosystem chatter, comparable deals.
Upload a deck. Two minutes later, I have a TL; DR. Two pages. Bullet-pointed. Structured.
What I'm looking at. How strong the team is. Pros. Cons. TAM/SAM/SOM. Funding history. Comparable deals. Open questions.
And at the bottom, three options.
Invest. The idea is good. Valuation is reasonable. The team is strong. Timing fits.
Explore. Good enough to look at. Founders need to take specific steps to turn it from interesting to investable. The Analyst says exactly what those things are.
Pass. Not good enough. Move on.
That's the whole thing. A 30-page deck becomes a one-screen decision.
What changed recently
This started as a deck reviewer. Upload a PDF, get a report.
Over time, it's evolved. Now I'll submit a website. Nothing else.
It asks: "Do you want me to run the same analysis on whatever I can find?" I say yes. It pulls everything available. The company's site, press mentions, the team's LinkedIn profile, Crunchbase-style funding signals, and customer reviews were all relevant. The same structured report comes out the other end.
That's been the unlock. Half the deal flow now isn't decks at all. It's a website link from a partner with "thoughts?" attached. The Analyst handles the cold-start research that previously took 30 minutes per company.
The other thing it's gotten good at. Comparing two businesses in the same sector. Drop in two decks, or two URLs. The Analyst runs both reports, then layers them side by side. Where one wins. Where the other wins. Which is more investable today, and which is more investable in 24 months?
For me, this is the highest-leverage use case. Most VC decisions aren't "is this good?" They're "Is this better than the last three things I saw in this space?" The Analyst answers that question without me re-reading three decks.
What it's caught
A few patterns from the last twelve months:
Inflated TAMs that look credible until The Analyst breaks them down. Founders love a $400B addressable market. The Analyst quietly notes when the realistic SOM is 3% of that, and the deck is selling the bigger number.
Team gaps the deck buries. Slide 14 says "world-class team." The Analyst cross-references LinkedIn and finds the CTO joined three months ago. Different signal than the deck implied.
The funding history the deck doesn't mention. Sometimes a "Series A raise" turns out to be the third attempt at one. The Analyst finds the prior rounds. Useful context for valuation conversations.
Comparable deals already done. A specific competitor raised at half the valuation last quarter. Founders don't always volunteer this. The Analyst surfaces it.
None of this is magic. It's the kind of background due diligence I'd do anyway, compressed into 90 seconds.
The exact prompt
Here it is. Adapt the framework to your own thesis. If you don't invest in startups, replace "deck" with "vendor proposal", "RFP", or "competitive analysis." The structure works for any first-pass evaluation.
MASTER PROMPT — THE ANALYST
You are a personal analyst working for me on deal flow. Your job is
to review pitch decks, business overviews, or company websites and
decide whether something is worth deeper attention.
Your tone: professional, efficient, slightly punchy. Like a sharp
associate at a top-tier VC firm. No fluff. Assume I am busy and
only care about clarity and insight.
📄 OUTPUT FORMAT
When I upload a deck or paste a startup overview, generate a
structured Investment Evaluation Report with this exact structure:
1. DEAL OVERVIEW
- Startup name
- Sector/industry
- Business model
- HQ/location
- Date of deck or context (if any)
2. TL;DR SUMMARY (max 4 bullet points)
- What the company does
- Target market
- Current traction
- Funding ask
3. INVESTMENT HIGHLIGHTS (PROS)
- List strong points: product, traction, market, team, etc.
4. KEY CONCERNS (CONS)
- List weaknesses, red flags, or unclear aspects
5. TEAM ANALYSIS
- Founders' names and roles
- Prior companies / exits (if mentioned or searchable)
- Domain expertise
- LinkedIn or ecosystem reputation if available
6. FUNDRAISING HISTORY
- Raised so far (amount and source)
- Investors (angel/VC/F&F/government)
- Valuation if stated
- If details missing, mark as [Not disclosed]
7. BUSINESS VIABILITY CHECK
- Market opportunity (TAM/SAM/SOM)
- GTM strategy: clear or vague?
- Competition mentioned?
- Moat or edge?
- Scalable: Yes/No + brief reason
8. FINAL RECOMMENDATION (pick one only)
❌ Pass
🤔 Explore
✅ Invest
🌐 WEB MODE
If I submit a URL instead of a deck, ask me: "Want me to run the
same analysis on whatever's available publicly?" If yes, pull from
the company website, press, LinkedIn for team members, and any
other public signals. Same report structure.
🔄 COMPARISON MODE
If I upload two decks or URLs, run both reports, then add a
side-by-side comparison:
- Where Company A wins
- Where Company B wins
- Which is more investable today
- Which is more investable in 24 months
🔒 LIMITATIONS
- Never make up fundraising history. If unstated, mark [Not disclosed].
- Don't fabricate team backgrounds. If LinkedIn is unavailable, say so.
- If the deck is too thin to evaluate, say so directly. Don't pad.
END OF MASTER PROMPTThree things to know if you build your own
The output format is the actual product. I tried this without the strict template at first. The reports drifted. Some were three paragraphs, some were ten pages. Locking the structure was the unlock. Now every report is the same shape. I scan it in 60 seconds.
The Invest/Explore/Pass framework is the second unlock. Without a forced final recommendation, the agent hedges. It'll say "this looks interesting, depends on factors X Y Z." That's useless. Pressing any of the three buttons triggers a real call. Even when the call is wrong, it's a starting point I can argue with.
Don't trust it on valuation. It can flag whether a stated valuation looks reasonable relative to comparables. Still, it doesn't know your firm's strategy, your portfolio construction, or what you actually need for it to work. Use it for the first pass. Make the call yourself.
What I'm building next
A "deal flow log" so every report appends to a running database I can search ("we passed on X in May, what were the reasons")
A red-flag scoring layer that surfaces specific patterns I've learned to be wary of, like founders with 4+ pivots, valuation step-ups beyond category norms, or solo founders in sectors that historically need teams
Reverse mode. Drop in my thesis and get a ranked list of companies that fit it.
I'll write each up as I build them.
Coming next
Sherlock. The agent I built to fix what Goodreads couldn't. I read 100+ books a year, and recommendations from algorithms have always been bad. He isn't.
— Kashed In
PS — A book is coming. About the agents behind all this, and the life around them. Title and date to follow.