Opportunity → AI For Normies
Framework → The 6 Company AI-Pilled Levels
Tool → Ontora
Trend → Customer Service’s Bumpy Road
Quote → No One Cares About Your Idea
Through the Anyhouck AI Fundraising Kit, founders get access to Notion fundraising agent workflows, including a 10,000-investor database, 50+ real pitch decks, and ready-to-use fundraising templates.
You’ll also unlock $15,000+ in founder credits and up to 3 months FREE of Notion’s top business plan to run your startup more efficiently while raising capital.
🔗 Houck’s Picks
My favorite finds of the week.
Fundraising
Pavel Prata on what seems to be driving VC fundraising (Link)
Growth
Adam Robinson on what will always be present when PMF is real (Link)
Chris Pisarski on the “email your boss” sales hack (Link)
ICYMI
An AI agent that handles customer conversations across calls, texts, chat, and email (Link)*
Internal Tech Emails showcases the texts between Sam Altman & Mira Murati - part of the evidence in the trial over Elon Musk's lawsuit against Altman and OpenAI (Link)
Bilal Farooqui shares two things that define a startup (Link)
Ryan Deiss on the graveyard for founders (Link)
Jordan Mitchell on how the founder “delusion” narrative is overplayed (Link)
Elizabeth Yin on the paradox of hiring (Link)
Opportunity: AI For Normies
It feels like everyone is using AI, right?
It feels like when you go home for the holidays this year, your grandpa will be telling you about his agentic team.
But the truth is we’re still so early.
Only 0.3% of the people in the world are actually paying anything to use LLMs.
Whether or not you agree with Andrew that IRL experiences are the best way to benefit from what AI has to offer, it’s clear there’s still so much whitespace for new ideas and new startups that leverage LLMs.
We’re not anywhere near the end of this run yet, even if OpenAI is starting to feel valuation pressure.
🧠 Framework: The 6 Company AI-Pilled Levels
Remember the mad rush around self-driving cars?
Waymo has largely solved it as a technical problem even if the service has been expanding outside of SF slowly.
You don’t hear much about the “levels of autonomy” for driving anymore.
But the concept of having different levels of AI-y-ness (yes that’s the technical term) are a good way to consider how companies have adopted LLMs across their entire business, too.
And Floodgate’s co-founder and GP Ann Miura-Ko did exactly that after hosting many companies claiming to be AI-native at Floodgate’s offices.
Her position is that most people are looking at whether they are “AI-pilled” as a binary idea when, in practice, companies differ both in how deeply AI is embedded as well as the technical capability of what AI is actually allowed to see, do, and change.
For example, a company where employees use ChatGPT to summarize meetings is not in the same category as one where agents can query systems of record, take bounded action, and improve the way future work gets done.
The questions Ann asked are:
What can AI see? How much context are you feeding it? Does the real way your company works only exist outside of queryable systems?
What can AI do? Summarization only or can it take action?
Who can extend the system? Engineers as the true power users only, or non-technical folks too?
How has the organization changed (relative to a few years ago)?
The whole piece is worth a read, but in short here are the six markers of what she considers to be a fully AI-native company:
🛠 Tool: Ontora
Context is everything (especially these days).
I talked recently about the importance of being an AI-native company with a queryable company brain, and a big part of setting that up is collecting an ever-increasing amount of context about what’s going on inside your company.
Ontora (YC-backed) is tackling that in an interesting way: their agents engage with and ask targeted questions to your team members about their work.
At first hand, this sounds like slowing them down, but the tradeoff is more/better context for the systems that keep your company moving faster across the board.
📈 Trend: Customer Service’s Bumpy Road
The bumpy road for the customer service sector since the pandemic continues.
a16z shared this chart on X:
And if you break look at a longer time horizon you see this:
No one has seemingly known what to do with customer service hiring since COVID hit.
Most famously, Klarna did a total about-face on AI-driven customer service:
In late 2024, Klarna’s CEO Sebastian Siemiatkowski announced that AI could already perform the jobs humans do. The company paused hiring for more than a year, cut its workforce from 5,500 to 3,400, and promoted an AI chatbot it said was performing the work of 700 customer service agents.
Six months later, customer satisfaction had fallen sharply, and service quality was inconsistent. Klarna was now asking software engineers, designers, and marketing staff to help answer customer inquiries. The same CEO who once minimized the role of people was acknowledging that the company had prioritized cost over experience.
The wrinkle that, now, hiring in the space is outpacing overall hiring hints that current sentiment towards AI could be that customer service is one of the areas least likely to results in layoffs in the longterm.
That may sound counter-intuitive, given how many CS-related startups there are pushing AI tooling, but it makes sense. When you have a problem with a product or service, who do you want to talk to? Surely not an AI. You want empathy, service, and the human touch then more than ever.
As we covered in yesterday’s deep dive with Cape, relationships are currency in a world where the busywork can be automated with LLMs.
💬 Quote: No One Cares About Your Idea
This tweet only got 25 likes.
A few years ago it might’ve gotten 100+.
On X these days it’s popular to talk about the rise of the idea guy.
Execution is perceived as easy (or at least less important than it used to be).
Let me tell you something.
Having the right idea has always mattered. And so has how you execute it.
Just because you can build software faster doesn’t mean building a big company is easier.
In fact, it might be harder.
If it’s easier to build something of your own, more people will want to do just that… leaving them unavailable to help you build yours.
Execution was never just about pushing code to production.