Sean’s AI Signal – Issue #49

There was no shortage of signals this week. Many focused on how expensive AI has become via the mainstream models (Claude, ChatGPT) and how companies are looking for ways to reduce AI spend. Many companies are turning to inexpensive and increasingly viable Chinese models. The result is companies are using more tokens and paying less at the same time. This is a signal worth watching.

But three essays I came across on X really made me think, not because they were about AI technology but because they all focused on what it actually takes to transform a company, an industry and a way of working in the middle of an AI revolution. I read them with a PeopleMetrics lens in mind. And about the insights industry, one that is already being disrupted by AI.

These three essays are this week’s issue. No charts, no model releases, no benchmark scores. Just three ideas that I think matter more right now than anything else I read this week.

Mitchell Hashimoto, founder of HashiCorp and creator of Terraform and Vault, posted a short essay on X this week that hit on something I ended with in Issue #48. You can read it here:

Mitchell Hashimoto@mitchellh

https://t.co/hXE5lXLpFT

12:26 AM · Jun 27, 2026 · 478K Views

111 Replies · 335 Reposts · 3.35K Likes

His argument is simple: taste is the ability to consistently make high-quality qualitative judgments where no objective metric exists. And in a world where AI is rapidly commoditizing production, taste is becoming the critical differentiator in all industries. A single person with a defined vision can now create what once required an entire team. But the ability to make qualitatively new judgments, to know what good looks like, remains distinctly human. I wrote in Issue #48 that taste and judgment were becoming the real differentiators in the AI economy. Hashimoto’s essay is the clearest articulation I have seen of why that is true and why it will remain true even as the models keep improving.

Why it matters: You do not need taste to produce a PowerPoint that creates a chart for every question in a study. You do not need taste to build an online dashboard that does the same thing. The insights industry has been doing both of those things for decades and AI makes them cheaper and faster than ever. But taste is what determines the story you decide to tell, how you tell it, how you bring it to life and how you make it endure beyond a single presentation. Taste is what separates a report from an experience, a finding from something a client actually feels and acts on. That is exactly what makes it a moat. You cannot hire for taste on a resume, you cannot mandate it in a job description and you cannot buy it with a software subscription. You know it when you see it. Any company, regardless of size, can build this moat. And that is super exciting to me!

Vas, CEO of Varick Agents, posted an essay on X this week that reframes the enterprise AI adoption problem. You can read it here:

vas@vasuman

https://t.co/LMqyFlorzC

10:04 PM · Jun 26, 2026 · 70.4K Views

24 Replies · 26 Reposts · 254 Likes

His core observation is that most enterprise AI fails not because the technology is wrong but because of how it is packaged. Companies hand employees a new interface, a copilot, a chatbot and then ask them to figure out where it fits into the work they have been doing for years. The AI does not own the workflow, the employee still does. And that is why adoption hovers around 25-30%.

Vas argues the fix is to stop building new places for people to use AI and start embedding it where the work already lives. Finance lives in NetSuite, sales lives in Salesforce, etc. The key point is not to replace those systems but to put AI inside them, running in the background, handling the manual work underneath the process, invisible to the people who benefit from it. In his words … “if enterprise software was the era of software as a service, enterprise AI is the era of outcomes as a service.”

Why it matters: The insights industry has been handing clients new interfaces for years. Log into this dashboard (and we will force you to change your password every 6 months). Download this deck (which might be too big of a file to actually open). Every one of those asks creates friction between the insight and the person who needs to act on it. The adoption problem Vas describes in enterprise AI is the same problem the insights industry has had with its own clients for decades, we just never talked about it this way.

The shift he is describing, from tools to outcomes, from interfaces to experiences that just happen, is exactly what we mean by bringing insights to life where the client already is working. A customer reel that arrives in someone’s inbox does not require them to log in anywhere. A short film about what customers are feeling does not need a training session. An interactive avatar of your customer segment does not live behind a password, it can join a zoom meeting. The insight reaches the person where they already are, in a Slack or Teams channel, in a form they can actually feel. On a Tuesday afternoon when they have 12 minutes between meetings and zero patience for logging onto yet another platform.

Most companies in the insights industry are still building new places for clients to consume research. The ones that figure out how to eliminate that step entirely, to just get the insight to the right person at the right moment in the right form, are going to look back and wonder why it took so long. PeopleMetrics is focusing on being one of those companies.

Jaya Gupta, investor at Foundation Capital and previously at McKinsey, posted an essay on X this week that I think is the most important of the three. You can read it here:

Jaya Gupta@JayaGup10

https://t.co/5vDrndcmFn

9:56 PM · May 8, 2026 · 4.06M Views

222 Replies · 629 Reposts · 5.18K Likes

Her argument is that in a world where products can be copied, categories can be renamed and technical advantages can collapse in months, the enduring question is what kind of organization you build around the people capable of building it? Everything visible gets imitated quickly. What’s more difficult to copy is the institution underneath. How a company attracts exceptional people, concentrates judgment, distributes authority and turns work into a compounding system no other company can reproduce. The shape of the company itself is becoming the moat.

She makes the case that the best companies are not just places where talented people go. They are structures that let a certain kind of talent finally express themselves. OpenAI was built around frontier model training as its gravitational center. Palantir was built around forward deployment and customer proximity as a status hierarchy. Neither fit the boxes that existed before them. Neither did the people who built them.

Why it matters: The people question is the hardest and most important one for me. Her essay triggered a question I have been asking myself lately: what kind of person are we actually building around and where do we find them? We have had a genuinely hard time finding people already inside the insights industry who are AI curious and ready to work in fundamentally different ways. The great news is the team we have is doing that work and doing it well. But it has led me to consider that maybe the answer is not finding insights professionals who are excited about AI. Maybe the answer is to find people who are AI curious by nature, who think differently about how work gets done, who are not yet attached to the way this industry has always operated … and then teach them everything they need to know about research. That is a different hiring philosophy than we have ever considered. It is also, I think, exactly the kind of shape change Gupta is describing. Not just new tools or new deliverables but a fundamentally different idea of who belongs here and what they can become.

I want to acknowledge something. This week I was writing about PeopleMetrics as much as anything else. If you are a regular reader, you know I try to write for a broad audience of business leaders. This week I wrote for myself as much as anyone.

Here is where we actually are. We believe taste is the moat but building an organization around taste requires the willingness to reject work that falls short of it … and that is harder than it sounds after 25 years of doing things a certain way. We believe the insights industry has been making clients work too hard to consume what they pay for and we are reimagining our delivery model around getting insights to people where they are, in forms they can actually feel, without asking them to log into anything or decode another dense presentation. And we believe the people question is the hardest one and that the answer may require us to look completely outside the insights industry for talent that is AI curious by nature and willing to learn this business from scratch.

None of that is finished and all of it is in constant motion. Some days it feels like we are ahead of something really important. Other days it feels like we are asking a 25 year old company to become something it has never been, with no guarantee that it will work.

But that is where we are. I hope that something in this issue landed for you too, whether you are reinventing a company, rethinking how you deliver your work or trying to figure out what kind of people belong in the next chapter of what you are building. These are not just PeopleMetrics questions. They are the questions of this moment.