First-party data beats the feeds you rent

Two publishers set the tone today, and they pulled in opposite directions. One laid out in public how it now runs growth through AI agents and the data it owns, while another firm's founder spent the day writing about racing in China and about friendship, with nothing to say about software at all.

First-party data and the signal tools it is meant to replace

Start with the term. A "signal tool" is a vendor that sells you third-party intent data: a feed that claims to tell you which companies are researching a problem like yours right now. You pay for the feed, your reps chase the accounts it flags, and you hope the timing is right. First-party data is the opposite. It is the information you collect yourself from your own interactions — who opened what, who replied, who booked a call, who churned and why.

SaaStr's latest edition of The Agents makes the case directly in its title: first-party data beats every signal tool [1]. The reasoning is not hard to follow once you separate the two. Third-party signal is a guess about strangers, sold to everyone who will pay for it, and often stale by the time it reaches you. First-party data is about people who have already touched your business. It is current, it is yours, and nobody else has the same copy. The trade-off is effort. Signal you can buy this afternoon; first-party data you have to capture, clean and keep. That is the unglamorous work of recording every interaction in one place and not letting it rot. We have written before about keeping customer data current, because a first-party advantage disappears the moment the records stop reflecting reality.

The session was run live, as a walk-through of the team's agent-led growth stack as it stands today [2]. That phrasing matters for a buyer. "As it stands today" is an admission that the stack is moving, not a finished product you can lift wholesale. Anyone selling you a fixed, final answer to agent-led growth is selling you something the people actually doing it will not claim.

What an AEO score is actually measuring

The same post reports a 46-of-50 AEO score [1]. AEO is answer engine optimization. For years the discipline was SEO: getting a page to rank in a list of blue links. AEO is the version for a world where buyers ask an AI assistant a question and read the answer it writes, rather than clicking through ten results. The question AEO tries to answer is narrower and harder. Not "does my page rank" but "when an answer engine explains this topic, does it cite me, and does it get my facts right."

A score out of fifty is one firm's way of making that measurable. Treat the exact number as less interesting than the idea behind it. If a growing share of your buyers now ask a machine before they ask a salesperson, then being legible to that machine becomes a distribution channel in its own right. It also means your public writing has a second audience that never used to exist: not just the human reader, but the model summarising you for them. The practical response is dull and durable. Write clearly, state facts plainly, and keep them consistent across every page, because an answer engine that finds contradictions will trust none of them.

Reading the people behind a tool

The other thread today was quieter and, for a buyer, worth a different kind of attention. When you adopt software you are not buying this week's feature list. You are committing to a company's judgement for as long as you keep the tool. One cheap way to read that judgement is to read what the people running the firm choose to publish when they are not selling anything.

Today the founder behind Basecamp did exactly that, twice, and neither piece was about software. One is a personal essay that opens with a trip to China in 2011 to go motor racing [3]. The other is about camaraderie — the argument that people should look after each other, inspire each other, and enjoy each other's company [4]. You can take or leave the content. The useful observation for a tool buyer is structural: a vendor confident in its product does not feel the need to manufacture a product announcement every day of the week.

There is a trade-off to name here too, so you do not misread it. A founder's personal essay is not a roadmap. It tells you something about temperament and values; it tells you nothing about what the product will do next quarter. The mistake is to treat a blog post about friendship or racing as a signal about where the software is heading. Keep the two separate. Judge the tool on the tool, and use the writing only to understand the people you are trusting with your data.

How to use all of this when you are choosing tools

Put the three items together and a working rule falls out. Prefer the data you own to the data you rent. Signal you buy is a commodity; everyone with a budget has the same feed. The record of your own customers, kept current, is the one advantage a competitor cannot purchase. That is why the first-party argument [1] is more than a slogan — it is a statement about which asset compounds over time while a rented feed resets to zero the day you stop paying.

Second, accept that discoverability is changing. AEO is not a reason to panic or to rewrite everything. It is a reason to be precise. The same clarity that makes a page useful to a person is what makes it quotable to a machine, and the firms measuring themselves against it [1] are simply admitting the audience has grown.

Third, weigh the vendor as well as the product. The essays published today [3][4] are a reminder that a software company is run by people with a worldview, and that worldview outlasts any single feature. When we have argued before about how to choose software worth using, the point was never the feature grid. It was whether the company behind the tool will make sensible decisions in the years you cannot see. On the same theme, being clear about what AI should and should not do in your business matters more as agents move from the demo into the daily workflow.

None of this requires a purchase today. It requires a habit: own your data, write plainly, and read the people before you read the pitch.