The Good Days Are Over
Investors stopped asking about your team. They’re asking about your data. Here’s what I’ve been sitting with since the book was finished.
These are notes I couldn’t fully articulate while I was inside the year with Arjun. Now that I’m standing outside it, I can name them clearly. I’m sharing them because I think some of you are living exactly what I lived, and because some of what I’m watching happen in the market right now is keeping me up at night in ways I didn’t expect.
There was a version of this that felt responsible.
You hired a data team and called it an investment. You approved an AI initiative and called it a strategy. You pointed to headcount, to tooling, to a roadmap, and the board nodded. Nobody asked what any of it was actually generating. The shelter wasn’t incompetent. It was the ambiguity that the market rewarded. As long as you were moving, the hard question stayed quiet.
I ran Stock Dropper that way. The decisions were rational within the context that existed. Investors asked about headcount. Boards evaluated on technical capability signals. The incentive structure pushed toward showing momentum, so we showed momentum.
None of it was building an asset. I didn’t have the language for that until the conversations around me started to change.
The conversation changed. Most of us missed it.
Two years ago, the question was how many data people you had and what stack you were running. Today, the question is how you’re maximizing the data you own to generate value, and what you’ve built internally that nobody else has.
That sounds like updated vocabulary for the same conversation. It isn’t.
It’s a financial conversation now, not a technical one. That’s why data has been moving from under the CTO to under the CFO in companies that are paying attention. Not as a reporting line preference, as a signal that data is being evaluated as a capital asset with expected yield, not a function with a headcount.
I decided that our work with Arjun, which I questioned at the time, I would assume direct ownership of the data portfolio myself. Not delegated. Not overseen. Owned. It created friction internally. People read it as a control move. It wasn’t. It was the only way I could be accountable for what the portfolio was actually returning versus what it was consuming.
When we got to the Series C conversation, I understood why that decision was the right one. The investors weren’t asking John about the data stack. They were asking me what our data was worth and how we were allocating toward it. I walked into that room equipped to answer. I wouldn’t have been able to say the same twelve months earlier.
Stock Dropper had 100 people when Arjun and I started. We have 200 now. That doubling didn’t happen because we threw more bodies at the problem. It happened because we finally understood which parts of the portfolio were generating yield and which parts were consuming capacity we needed elsewhere. The operating model changed first. The growth followed.
What the good days actually cost
Every team is designed to show intent, and every roadmap aims to demonstrate momentum, so they created something. However, it wasn't the value I expected. Instead, complexity arose: a portfolio of products, pipelines, and processes stands between me and the answer needed for the new investor conversation.
The cost wasn’t just what I spent. It was the distance between where I was and where the conversation now needed me to be. And that distance doesn’t show on a dashboard. It shows up in the fundraiser. It shows up when someone asks what your data is worth and you reach for a headcount number instead of a yield number.
I lost time. I’m naming that plainly because I think some of you are losing it right now, not because you’re making bad decisions, but because you’re making rational decisions inside an incentive structure that has already shifted around you.
The operating model that no longer fits
Initiatives made sense in a linear world. You plan a year, execute, measure, plan again. Clean, defensible, boardroom-friendly.
We are not in a linear world.
The companies generating real yield from their data aren’t asking what they’re building this year. They’re asking what is generating value today, how to allocate toward it, and what to kill to protect the capacity to build what they’ll own tomorrow. That is a portfolio management question, not a roadmap question. It doesn’t fit within an annual initiative structure. It requires constant reallocation based on what’s actually returning versus what’s consuming resources and generating noise.
That shift is not a technology decision. It is not a team structure decision. It is a CEO decision. And most CEOs I speak to haven’t made it yet, not because they don’t see the need, but because the initiative model is what their boards were trained to evaluate.
The agent questions that nobody is answering honestly
This is where I want to be careful. Because this is where I see the most risk and the least honest accounting.
I know CEOs who have built what I’d call agent armies. Dozens of AI agents are replacing functions, automating workflows, and running processes that used to require people. Some of them announced it as a transformation. Some quietly restructured headcount alongside it. From the outside, it looks like a clean trade, fewer people, more automation, lower cost.
But when I ask about the actual cost of running those agents, now, and in two years as usage scales and model pricing shift, most of them can’t give me a number. Not a rough one. Not a directional one. The headcount line got smaller. The infrastructure cost is sitting somewhere in engineering budgets without a name on it. And the return, the actual yield from what the agents are producing versus what the people produced, nobody has run that comparison cleanly.
We are in danger of repeating the exact mistake the good days trained us to make. We replaced one thing we couldn’t measure with another thing we can’t measure, and called it progress because the org chart got flatter.
I’m not saying agents are wrong. I’m saying the question of what they actually cost to run, what they’ll cost as the underlying model pricing changes, and what they’re genuinely returning versus what they’re just doing, that question is not being asked with enough rigor. And the CEOs I see moving fastest on this are the ones who will have the hardest conversation with their boards in eighteen months when the numbers finally surface.
The IP reality and what it means
Running on generic models is not a competitive moat. Every company has access to the same models. Which means the differentiation has to come from what you build on top, the proprietary layer, the internal dataset, the model trained on your own operations that cannot be replicated from outside.
Most companies are somewhere in the middle right now. Some generic model usage for productivity, some internal tooling being built, some data work that is genuinely proprietary, and some that isn’t. The question is which parts of your AI investment are building toward something you’ll own, and which parts are operational costs you’re calling innovation.
I’m in the middle of this myself. Some of what I’m looking at changing built reputations inside my own team. Some of what I need to protect doesn’t yet have a strong internal advocate. The companies that will matter in three years are the ones that can point to something proprietary, a dataset, a model, a system, that becomes the asset an investor can value. Not the number of agents running. Not the tools integrated. The thing you own that nobody else can replicate.
Your data is that thing. If you’re not treating it as your primary IP, if it’s still sitting as an input to operations rather than an asset being actively managed, the investor conversation will eventually surface that gap. Better to surface it yourself.
What I ask myself every quarter now
Not a framework. Just what I actually do.
Which of our current data products can be connected to a revenue or cost number without explanation?
Which AI investments are building something we will own in two years, versus something we’re consuming from a vendor whose pricing we don’t control?
What are we maintaining that we would not rebuild if it broke tomorrow?
What came back from those questions told me more about where we actually stood than years of roadmap reviews. Some of the answers were uncomfortable. The uncomfortable ones were the ones worth knowing.
Do you see what I see?
The good days sheltered us from activity. Hiring as proof. Initiatives as momentum. AI adoption as modernity. Each stage felt like progress. None of it was compounding.
What I see now, and what keeps me up at night, is that we may be building the next version of the same mistake. Agent armies with unquantified running costs. AI initiatives without a proprietary layer underneath them. Operating models that got flatter without getting clearer about what they’re actually returning.
The investors are asking the right questions now. The pressure is real, and it isn’t going away. But the risk I see in the companies moving fastest is that they’re optimizing for the investor conversation rather than for the underlying reality, and those two things are not the same.
I don’t have a clean answer. I have 200 people, a portfolio I can finally see clearly, and a set of bets I’m making with genuine uncertainty about which ones are right.
What I know is that the dissonance is not over. The shift is happening, and none of us gets to watch from the sidelines. The question is whether you’re moving toward something you’ll own or toward something that looks right until the numbers come in.
That’s what I’m sitting with. If you’re sitting with the same thing, tell me what you’re seeing. I’m thinking about what comes next, and I’d rather figure it out in conversation than alone.
Two hundred people. Double what we started with. The operating model changed first. I’m still watching to see what the growth does next.