👋 Hi, I’m Andre and welcome to my newsletter Data Driven VC which is all about becoming a better investor with data and AI.
ICYMI, check out some of our most read episodes:
Brought to you by Vessel - Agentic fund operations for VC and PE firms
The problem isn't that you need better software. It's that your fragmented tech stack is killing your firm's productivity.
Vessel is the agentic operating system that unifies your fund accounting, portfolio data, and LP records, so AI agents can draft quarterly reports, prep LP meeting briefs, and answer LP questions with full context.
You focus on relationships. Vessel handles the operational gruntwork.
Earlier this week, we hosted another DDVC x SuperReturn breakfast in Berlin.
150+ applicants. 30 select participants. 2 hours. 1 room. Deep conversations.
Here’s a summary of my key takeaways👇
Finding the edge before everyone else
Spotting stealth companies. The best founders build quietly before they announce, and AI now makes that quiet phase visible.
The signals are everywhere if you instrument for them: new domain registrations, hiring for roles that only exist after a thesis is set, repeat operators making sudden moves, early commit and product traces.
Lots of this data has become broadly available via providers such as Harmonic, Evertrace, and others. There’s little value in rebuilding the scraping and pipelines. Focus should be on scoring to match your fund focus with highest probability of success opportunities.
Network intelligence. Your warmest path to almost any founder already sits inside your team's collective inbox, calendar, and relationship graph, mostly unused.
The funds pulling ahead treat their network as a queryable asset, not a memory locked in one partner's head.
Map who knows whom across the whole firm, then let an agent surface the strongest intro path on demand. Relationship capital only compounds when it is searchable. Many funds use Affinity as a baseline.
Leverage your knowledge base as a second brain. Every call note, memo, and passed deal is training data for your own judgment, and most of it rots in folders.
A well-structured second brain lets you ask "what did we conclude about this category two years ago" and get a real answer in seconds.
The firm that remembers its own thinking makes faster and more consistent decisions. Invest in capture and retrieval before anything flashier.
Building the stack
Claude and Codex have become the exclusive interfaces. You don’t need fancy dashboards anymore. Off-the-shelf solutions are the lowest-risk, highest-frequency win, and the room agreed it is where most funds should start.
Drafting LP updates, memos, market maps, and outbound at near-zero marginal cost frees partners for the judgment work only they can do.
Map your workflow into skills, share it with your team, find best practices, and institutionalize your processes and judgment.
Loops and agents. A single prompt is a tool. A loop that runs, checks its own output, and retries is the beginning of an employee.
The shift this year is from AI that answers to AI that does, chaining steps across sourcing, enrichment, and follow-up without a human in every link.
Start designing workflows as loops, not one-off queries. The compounding value lives in the steps you no longer touch. Here’s a great point to start your journey.
Platform versus use-case-specific. Horizontal platforms give you flexibility and scale. Vertical tools give you depth out of the box.
The honest answer is sequencing: start narrow, define a clear goal, and deliver tangible ROI. Then scale from there based on a unified foundation.
Build versus buy. Building feels like control and often becomes a maintenance burden no one at a fund is staffed to carry.
The pragmatic line: buy the commodity, build only where your edge is genuinely proprietary, such as your exclusive first-party data or scoring logic.
Every internal tool you build is a tool you must maintain. Be honest about whether that is the best use of a small team.
Reliability. The fastest way to kill internal adoption is an agent that is right most of the time and confidently wrong the rest.
Trust is the real adoption metric, and it is earned through guardrails, human checkpoints, and narrow, well-scoped tasks before broad ones.
Design for reliability before reach. One embarrassing hallucination in front of an LP costs more than months of efficiency gains.
Join 1693+ investors in our free Slack group as we automate our VC job end-to-end with AI. Live experiment. Full transparency.
Bringing the team along
Training and the cultural shift. Tools do not transform a fund. Behavior does, and behavior change is slower and messier than any software rollout.
Unfortunately, everyone in the room agreed that this is the biggest bottleneck. Still.
The funds making real progress treat this as change management, with internal champions, shared wins, and explicit permission to experiment. A best-practice seem internal hackathons or co-development sessions.
Budget as much energy for adoption as for the technology itself. The bottleneck is almost always cultural, not technical.
Rolling out skills. Capability spreads fastest when it is packaged and shared, not rediscovered by each person from scratch.
Codifying a good workflow into a reusable skill the whole team can run turns one person's breakthrough into the firm's baseline.
Capture and distribute what works. A skill used by ten people is worth ten times the same skill trapped with its author.
VC operations. The back office is where AI pays off quietly and immediately: portfolio monitoring, reporting, data rooms, scheduling, CRM hygiene.
It is less glamorous than sourcing the next breakout, and the return is far more certain.
Do not overlook operations in the rush toward sourcing magic. Clean, automated ops free your best people for the work that actually needs them.
Experiments versus scaling. Most funds are stuck at the demo stage, with impressive one-off experiments that never become daily habits.
The gap between a cool prototype and a scaled workflow is process, ownership, and reliability, not more clever prompts.
Pick one experiment and push it all the way into production before starting the next.
One scaled workflow beats ten abandoned pilots.
The most frequently named challenge these days
Compliance constraints. This is where enthusiasm meets reality, and it is non-negotiable.
Most funds in the room were regulated in one form or another. Based in Europe? Nice, how about GDPR, European AI Act,…?
Confidential data means you cannot pipe everything into a public model without thinking hard about data handling.
Learning? Someone needs to wear the compliance head. Bring this person into the room early, as a design constraint rather than a blocker.
If you get stuck, pull a GP / MD in to resolve grey areas. Either take the risk or not, but someone needs to decide so you can move on. Document the process, reuse, and improve.
The funds that get this right build trust, and the ones that do not will eventually pay for it.
The throughline
The clearest signal from the table: the technology is no longer the hard part.
The edge in 2026 belongs to the funds that change how their people work, scale the few things that work, and respect the constraints that come with managing other people's money and data.
We concluded the round with everyone sharing their top tool (Claude, Codex, et. al not allowed to mention). Here are the 5 most frequently mentioned:
Thanks to Roundtable for partnering on the breakfast, and to everyone who showed up in the midst of the SuperReturn rush for an honest and open discussion.
Join our next DDVC x Bits & Pretzels Breakfast 30th Sept in Munich.
We will keep sharing what we learn.
Stay driven,
Andre
PS: Check out Vessel to automate your fund operations
