What’s 🔥 in Enterprise IT/VC #454

Striving for the autonomous enterprise + and what we're excited for with our new $250M inception fund!

We all know agents and GenAI are the future, but having lived through so many cycles from the .com era to cloud and mobile, I remind myself that AI is just a technology. With transformative tech, we always need to think about why is this tech going to help us create new things that could not exist before or how can it help us reimagine existing solutions. To that end, we announced boldstart ventures Fund VII, our new $250 Million fund, to back technical founders from Inception who are building the “autonomous enterprise.”

👇🏻 to watch the video and read the full 🧵.

Here’s a bit more about what the “autonomous enterprise” means. The world is never getting to a fully autonomous or self-driving business but the aspiration is what counts and striving for the future is what gets us excited. That being said, we have a long way to go and you see, I never said we’re an AI fund 🤣.

The Generalist AI video below (amazing team btw with 2 co-founders from Google Depemind and other from Boston Dynamics)

As readers of this newsletter I know you know what we mean by “Inception Investing” but here it is for the newer folks:

Finally, it’s been quite a journey. We got our ass handed to us so many times, had so many rejections, and just 🙏🏻 to be here still standing, doing what we love ❤ every single day - partnering with founders who see and are building the future and for us just to be a small part of that journey!

From a $1M fund in 2010 to over $1.1B of assets under management (AUM) today, the journey has exceeded anything we imagined. While AUM isn’t our north star, it gives us the firepower to double down on our core mission: helping founders build, wire, and secure the autonomous enterprise. With our largest fund to date at $250M, our lead initial checks range in size from $500K to $15M. And as you scale, we can continue investing out of our $175M Opportunities III fund which still has plenty of dry powder.

read our full blog post here!

One last thing - yes, I conceived of and created the video of the Yeti and Bigfoot myself, 8 seconds at a time, using Google Flow. Here’s the prompt that kicked off my deep dive into creating and producing my first AI generated video.

Create a realistic, entertaining YouTube vlog video in the style of the channel "Outdoor Boys." The main characters are a friendly, large Yeti with shaggy white fur and Bigfoot from the channel "BigfootVlogs," both standing in a startup type office, drawing on a whiteboard. The mood is friendly, energetic, and authentic, just like Outdoor Boys. The video should look like a genuine, spontaneous scene from a real vlog, not cinematic or overly polished—just natural, handheld, and authentic.

Yeti and Bigfoot are drawing by a whiteboard." They are drawing technical diagrams for their new startup. Yeti grins at the camera and says: "Pre seed, seed, post seed - what the F! I just need someone who'll lead my first round on this alone!" Bigfoot steps away from whiteboard, thinks, and then says: I know. (slight pause) Inception! (emphatically) We need boldstart!"

Camera style: POV, selfie stick, occasional drone shots capturing the gritty startup office and the playful chaos, but always keeping the raw, vlog-like feel—unsteady, handheld, and personal. Yeti personality: Humorous, bold, likable, practical. Props: whiteboard, markets, but otherwise what's in an office. have small boldstart logo from website https://boldstart.vc on bottom of video like a watermark

Try it out here - it’s so much fun!

And here’s another fun BigFoot video launched hours after hours with 2m views!

Based on spending a dozen hours or more making our launch video, I do 100% agree with Ethan’s comments 👇🏻 - only way to learn is to just dive in and have fun!

As always, 🙏🏼 for reading and please share with your friends and colleagues!

---

Scaling Startups

#stay the course - need your own center of gravity - i call this advisor whiplash

#along the same lines - don’t overindex just to make sure people think your idea is bad because many are, but…getting that headshake you’re crazy kind of look from someone is just right, usually folks will say it can’t be done, market too early, etc…

#inspiration from founder/ceo of Perplexity - keep building!

#and need more of this 👇🏻

#pretty awesome post on your marketing team as all agents!

#ARR at Series A 📈

---

Enterprise Tech

#wowza - that Microsoft deal may make it much harder in future for OpenAI to make any big AI type acquisitions - Windsurf is no longer going to OpenAI for $3B, now Google paying $2.4B for “licensing” fee and hire its CEO and some key employees (The Infoprmation)

#🎯

#definitely worth watching and some nuggets of wisdom

#browser wars heating up again with Perplexity’s browser Comet out and ready to automate some of your most common tasks

#🤯 if true, wowzah, gaining market share rapidly vs. cursor and others…

#any one can code a prototype but can you build a business? distribution matters more than ever!

#is AI Security a market or not? From The Information - For Startups Safeguarding AI Models, Revenue Remains Modest (The Information)

#speaking of distribution, cluely looks like a way better Granola but what they have gotten right is marketing and distribution - bold attempt to try something different

#agents shopping and buying on your behalf is definitely one of the first agentic use case to roll out so this is a great listen/watch to better understand

#didn’t take long for the WSJ to bring together “vibe coding” and “enterprise” in a story but as Choice Hotels CIO says: “But for vibe coding to really take off, it needs to work inside of a company’s existing code base and connect to its other core development systems.”

#totally agree, everyone wants agentic processes but still so early…

#😮‍💨 sign of the times? What the hell is going on???

#glass half full or half empty? Research paper from Carnegie Mellon showing agents only completed 30% of tasks - IMO, as the models and tools get better, this will dramatically improve! (paper here along with my summary)

🧠 Can your AI agent actually do the work?

A new benchmark from CMU just dropped—TheAgentCompany—and it puts today’s top LLM agents in a simulated SaaS startup to test their skills. The results are… revealing.

👉 175 tasks across engineering, HR, admin, PM, finance
👉 Real tools: GitLab, RocketChat, OwnCloud, Plane
👉 Agents browse, write code, chat with teammates, submit reports

The top performer? Gemini 2.5 Pro—which only managed to complete 30% of tasks autonomously. Even tasks like reimbursing expenses or summarizing project metrics proved too difficult for most agents.

Why does this matter?

💡 Most LLMs are great at code. They struggle with:
	•	Long-horizon workflows
	•	Social reasoning (e.g. “follow up with Sarah”)
	•	Navigating messy UI environments
	•	Real-world ambiguity and fuzziness

🛠️ The hardest part of agents isn’t intelligence—it’s orchestration.

Autonomy requires memory, context, planning, and adaptability. And right now? We’re still early.

But benchmarks like TheAgentCompany show us where the gaps are—and where the next billion-dollar startups will emerge.

🔥 Want to build in this space? Start with:
	•	Workflow loops humans hate
	•	High-context, high-repeat tasks
	•	Narrow domains where partial autonomy is still valuable

Agents won’t replace humans anytime soon—but they’re coming for the busywork. And that’s a huge opportunity.

#

---

Markets

#🤯

#all your 🦄 in one infographic from CB Insights - also some interesting data on the new batch (congrats Clay with one of highest revenue/employee at $1m)

1 in 5 new unicorns are AI agents, with AI taking over the unicorn landscape, representing 53% of all new billion-dollar companies in 2025 so far. Among the newest unicorns, 12 are building AI agents, including Hippocratic AI (healthcare), Cyberhaven (data security), and Parloa (customer support). 

Newer unicorns generate 83% more revenue per employee than older ones, with $814K per employee on average, compared to the $446K average across all unicorns. This reflects automation-first approaches and leaner operations that avoid the operational bloat older unicorns accumulated during their growth phases. For example, among unicorns born in 2025, the company with the highest revenue per employee is soft drink company Olipop ($1.2M/employee), followed by AI sales agent unicorn Clay ($1M/employee).

#