What’s 🔥 in Enterprise IT/VC #442

The Enterprise Agent Hunger Games - the agent wave 🌊 is just starting—but without the last mile of IT infra/security it’s all vibes and no production. 🧠⚙️

While the whiplash we are experiencing from our current administration can feel nauseating at times 🎢, it is quite clear that the net effect will be a massive accceleration for AI and agentic adoption in the enterprise, for pilots moving into production, for agents to eat into labor.

Which is why we have Accenture

and why the largest of Fortune 20 companies are discussing progress with AI on their earnings calls all the time.

Here’s Aaron from Box chiming in yesterday

This is what I also see speaking to many IT execs from the F500.

Which is why I wrote this short LinkedIn article on Wednesday this past week to help synthesize some of my own thinking - every company is agentic now… (excerpt in full below)

The Enterprise Agent Hunger Games

April 16, 2025

This started out as a post but too long so now an article - just sharing some initial thoughts on agents that actually do work in the enterprise...with roles/tasks/plan/reason/take action/learn/rinse and repeat

Met founders this past week from all over the world - common theme - every company is an agentic company, all agents all the time. Very little differentiation on tech stack, pick a framework, pick a model...

However, most of these founders should NOT be raising venture capital - many founders can build nice, lean lifestyle businesses from their agents.

If one chooses to go down the VC path, know what you are signing up for - whatever you build, the only question is how big can it get if everything goes right.

And if you're going in the enterprise, the demands are extremely high - you can't vibe code your way into production - that last mile is everything - security, privacy, compliance, integrations, determinism and repeatability, accuracy, memory, cost, orchestrating swarms of agents, and even more security

When it comes to large enterprise deployments of agents, there will NOT be a thousand point vendors running amok in an enterprise infrastructure and nor will large organizations rush to rip and replace their systems of record

So path of least resistance will be buying off the shelf agents from existing SaaS vendors to leverage existing sunk costs and investment.

Next, net new AI native startups may be able to wedge in for specific workflows, say security, for example, which must sit on more data than just one incumbents platform like a Palo Alto or Crowdstrike, perhaps getting data from existing SaaS apps and other areas to provide contextual intelligence and be an agent/workflow sitting on top of multiple systems of record. Think about this for sales, success, etc - need to be able to provide intelligence beyond just one system of record to deliver value beyond off the shelf.

Sell that initial product with quick time to value, market a bigger vision so enterprises know you won't just be another point product - hard to do but possible. Your team's credibility and experience is super important here as these enterprises bet on the jockey especially in the early sales.

And enterprises are keen to try everything now, dispose of most, then make their decisions which is why most AI is still in pilot phase now versus full deployment

Think about the moat as well over time - sure it's time to value, shipping really fast, but the data matters and part of that comes from having more and more agents doing more and more work and learning from each interaction.

Finally, there will be lots of custom built agents in the enterprise and many who start with an initial wedge will vie to be one of the main platforms from which all of the internal agents are built. This is a massive battleground where many are entering from so many different angles from the cloud providers, to the foundational model providers, to the AI native app cos like Glean, Writer , new frameworks for multi-agent like boldstart ventures port co CrewAI , existing cos like Zapier and of course platform cos like ServiceNow and Palantir Technologies - going to be an interesting battle.

This is how venture scale returns will be created as thesse valuations for these private startups escalate quickly - IMO, not all will be able to grow into their numbers and many will go away with a few creating ginormous value. This is the game on the field now.

My colleague Ron Miller also covered of this in his Forward Thinking commentary from his amazing weekly newsletter FastForward (Ron was a 10 year enterprise reporter at TechCrunch so I highly recommend subscribing!)

It’s easy to succumb to the hype, but Thomas Squeo, Americas CTO at tech consultancy ThoughtWorks, says fundamentals still apply with agents. “At the end of the day, you have to manage it as a piece of enterprise software because vibe coding alone (coding with the help of AI agents) is only going to get you in trouble,” he said. Even with agents in place, maybe especially with them in place, large companies still need to pay close attention to how they manage risk.

“I think that the biggest issue that I've seen is that enterprise risk is the force of gravity that everybody has to pay attention to,” he said. “That’s because as soon as you have to disclose how you're maintaining a production system, and what the risk profile of that system is in the enterprise, especially for publicly-held companies, that disclosure risk makes everybody more conservative.”

By the way, if it’s not obvious, I’m super pumped about the idea and need for a whole complete infrastructure that needs to be made and reimagined as Agentic traffic grows to 10-100x human traffic. Here’s one of our latest investments Agentuity explaining why this needs to happen!

And the developers at React Miami also believe we need a new agent native cloud! Congrats Agentuity!

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

---

Scaling Startups

#💯

#In the world of AI, everyone seeking PhDs but are they seeking the right ones?

Poor, hungry, determined

PhDs for the win

#❤️ this - super powerful - btw, this is what Agentuity is doing - fun to watch the progress (week 9 video here as they create a Heroku-like platform for agents)

#don’t believe everything you see on social - yes vibe coding is great but still there is so much under the hood to build enterprise class software and also the maintenance, security headaches etc. but, we will no doubt have more simple apps built easily by non-devs for simpler use cases

#sign of the times…🤦🏼‍♂️

#👇🏼 - we need some originality folks!

#when it’s this easy to ship a MVP, distribution and taste becomes an even bigger moat along with the team

---

Enterprise Tech

#and so it begins - amazing how Windsurf outran Github CoPilot as its initial wedge in the enterprise - the speed of shipping and updating vs. a large incumbent who had the data was insane. The real question though is if Windsurf users stick with OpenAI as currently it used Anthropic Claude beneath the hood. Sure there are switching costs from the UI, but many folks I know use both Windsurf and Cursor already. And oh yeah, CNBC broke the news that Windsurf was the second choice as Cursor turned them down

#fun discussion in comments (click through 👇🏼)

#agree or disagree - lots of comments 👇🏼

#why startups always have a chance even over distribution and data moats

#🤯📈

#🎯 why I’m 🔥 up about Tessl, a port co

#just a reminder - all this in 2 years, what will the next 2 years bring us? now apply this to lots of other areas where AI can eat the world…

#this is pretty cool - try it out - upload an image and ask it a question, to reason…

#we are now just upleveling the idea of Forward Deployed Engineers (FDE) - get ready for ludicrous speed!

#great post on pricing for agents - 4 current frameworks from Manny Medina, founder of Outreach and now newco Paid on Kyle Poyar’s newsletter Growth Unhinged - IMO lots of hybrid versions of this now being explored…btw, the price per action definitely the future but customers need some predictability so I would suggest bands or buckets to help customers realize this. sellling FTE replacements one needs to be careful as if you’re buyer is a manager who cares about their empire or headcount, or if a decision maker is worried about being fired, then well, you may have an issue…

#the continued challenges scaling AI = massive opportunity ahead

#one of the most powerful tools for hackers growing virally in the Darknet - from SlashNext (👇🏻 below)

#there is no AI in the enterprise without AI Security!

#well Lovable now legitimized if it weren’t already so esp. as Figma files for an IPO!

#Fortune 500 AI deployments - where we are in the cycle - from “12 Graphs That Explain the State of AI in 2025” - Stanford's AI Index tracks performance, investment, public opinion, and more

#autonomous software engineers from OpenAI are coming - link to video in 🧵

#how GPT 4.5 was built

---

Markets

#simply insane the growth of VC and private assets in general - huge implications - what is venture capital?

#due to all of this private capital, there is less of a need for companies to go public as there is a liquid market but only for the top 20 cos - here’s anecdotal data on the rise of secondaries for these top tier later stage companies from the Pitchbook Q1 2025 VC Monitor.

#not just VCs sitting on portfolio companies and needing some liquidity