This blog post is based on a talk I did recently at DDVC Conference. This was originally generated by AI based on the talk transcript, which I edited.
At Untapped, we’re constantly experimenting with AI to supercharge how we operate — from sourcing deals to supporting portfolio companies. Over the past year, I’ve tested dozens of tools, built a bunch of scrappy ones myself, and learned a lot along the way.
In this post, I’ll walk through how I imagine the future of VC, the tools we’ve tried, and what’s actually worked. Whether you’re new to AI or already building your own automations, my hope is you’ll walk away with at least one idea to try out.
👀 The Vision: What AI-Powered VC Looks Like
Here’s what I imagine an AI-first VC firm looks like:
- General Tasks: AI that understands your goals, calendar, and to-do list — and just handles stuff for you.
- Networking: Dynamic relationship maps, smart intro suggestions, and event recommendations based on who’s going.
- Deal Sourcing: Intelligent alerts, social signal tracking, and predictive scoring.
- Due Diligence: Auto-research companies, LLM-powered SWOTs, and deep contextual analysis.
- Portfolio Support: Automated monitoring, resource matching, and reminders based on portfolio updates.
- Admin & Reporting: Legal docs and spreadsheets turned into simple dashboards. Auto-generated LP updates.
- Fundraising: Prioritized LP follow-ups and intelligent pipeline management.
It’s a big dream, but we’re building towards it — one small experiment at a time.
🤝 Networking: My Favorite Hacks
- Interaction Summaries: Every email I get is summarized and tagged to the sender — so before a call, I can quickly skim our entire history in seconds. It rolls up to the domain level too. Huge time-saver.
- Auto-Intro Suggestions: Tag a founder “healthcare” → see a list of healthcare investors pop up on their profile in real-time.
- Happenstance: Natural language search across LinkedIn, Twitter, and Gmail. Even better — you can “friend” someone and search their network too.
- Bridge: Easiest way to make intros. One-click, fully tracked, sent from your email.
- Fathom: Meeting note-taker that auto-generates summaries and action items.
- Custom CRM*: Currently building my own AI-native CRM that pulls in email, calendar, notes, and tasks into one view. Still early (two weekends in), but the goal is seamless context and automation across everything.
🔍 Deal Sourcing: Let AI Do the Scanning
- Mean VC (Custom GPT): My favorite GPT. No matter what you pitch it, it’ll challenge your assumptions like the world’s grumpiest partner. Founders love it too — we use it internally before pitch reviews.
- DealFlow Digest: Founders apply, investors sign up, and we match them automatically based on tags.
- Deals@untapped: I forward deals to this internal email, it extracts key info and updates our pipeline. No forms needed, thanks to LLMs.
🧠 Due Diligence: Faster, Deeper, Smarter
We built a custom AI tool that:
- Finds similar startups from our CRM and unicorn database
- Summarizes reviews (Product Hunt, BuiltWith, etc.)
- Does SWOT analysis, customer pain point breakdowns, and more
Eventually, we found a better solution — Wokelo. It generates 20–40 page research reports across private and public data sources. Expensive, but powerful.
Bonus: Their biggest customer (a $2.5B fund) scaled from 150 → 800 diligence reports/month just by switching to Wokello’s API.
📬 Portfolio Support
- Updates@untapped: Forward a company update email → extracts metrics, burn, challenges, and drops them into Airtable.
- Email Relationship Analyzer: Founder sends a list of 150 VCs → my tool analyzes past email volume and spits out who we know best, by firm and by person. Took 15 minutes to build, now I use it constantly. (link)
💸 Fundraising Prioritization
We dropped 150 LPs from our pipeline with quick notes into ChatGPT, asked it to sort by last touchpoint, and tell us who to ping this week. Shockingly effective. We used to do this manually.
Other tools:
- Ottogrid: AI agents inside a spreadsheet. Great for enrichment at scale.
- ChatGPT + Airtable: We built simple queries into our CRM for LP outreach, update history, and follow-up suggestions.
🚀 Getting Started with AI as a VC
Beginner:
- Use ChatGPT and Perplexity daily
- Try Zapier for basic automations
Intermediate:
- Build scrappy tools with Replit Agent or Bolt.new
- Try AI CRMs like Atio, Folk, Hearth
Advanced:
- Connect your email, notes, calendar, and CRM
- Build an end-to-end AI copilot (I’m trying!)
🧪 Build Culture > Tools
The biggest unlock? Build culture.
My team now tries tools on their own. We do “build-with-me” hours where I build tools in front of them. They help me maintain them now.
The #1 advice I’d give to firms: Let whoever’s building shadow the GP for a week. Tools only work when they match your actual behavior.
Final Thoughts
I’ve built 100+ tools (see them at yohei.me). Some lasted a day, some I use daily. Many died when models deprecated. I don’t fix all of them — I just keep building.
The cost of experimentation has dropped dramatically. A few prompts and a weekend is all it takes. Even if the tool doesn’t stick, I always learn something new.
If there’s one thing I hope you take away from this post: experiment. Play around. AI’s moving fast, and the best way to understand it is to get your hands dirty.
Leave a Reply