The dominant story about AI is still productivity, but it's shifting to queryability. When an organization can read and synthesize what it knows, decides, and learns, it stops behaving like a collection of departments and starts behaving like a single intelligence that improves over time.
Right now, the dominant story about AI is still a story about productivity, with faster engineers, faster marketers, and copilots everywhere. That story is not necessarily wrong, but it is incomplete, and the more interesting shift I’m beginning to see is from productivity to queryability.
When the things your organization knows, decides, and learns can be read and synthesized, it stops behaving like a collection of departments and starts behaving like a single intelligence that improves over time.
Diana Hu of Y Combinator calls this the queryable organization, and she believes incumbents will struggle with it more than startups will. She's right, but I think it's ultimately a solvable problem. I would also argue that it has to be solvable. A mid-market company has twenty or thirty years of proprietary signal no startup has: customer behavior, deal patterns, operating wisdom, hard-won judgment. Startups have to build that signal from scratch, but incumbents are sitting on it, even if most of them don’t know what to do with it.
My contention here is that the real opportunity is not for startups. It is for the companies that already have the signal and have not figured out how to close the loop.
Open-Loop is the Default
Walk into any mid-market organization and the pattern is the same. Every day the sales team understands more about why customers buy. Every day the customer success team learns more about why they leave. Every day the operations team learns more about what the business does well and where it struggles. Every function is generating an enormous amount of intelligence. The problem is that none of it is becoming organizational intelligence. None of it is being socialized or operationalized, which means most of it is useless outside the department it came from.
A customer success representative finishes a call and gets ready for the next one. She does not surface what she learned, not because she is opposed to doing so, but because no one is asking for it and there is nowhere for it to go. Marketing keeps producing the same collateral. Product keeps prioritizing the same roadmap. The same lessons get relearned next quarter, and the quarter after that.
This is what running open-loop looks like, and it is expensive in a way most CEOs cannot see. It is also hard to quantify, because the cost is the absence of compounding rather than the presence of a problem.
The Closed-Loop Organization
The way to fix this is to design a closed-loop organization, one that runs a disciplined practice that turns observation into learning, and learning into structural change.
It is worth thinking about this in two parts: where signal comes from, and how it gets processed.
Where signal comes from
There are four dimensions of signal:
Customer intelligence: what customers are telling us, directly or by behavior.
Operational intelligence: what the business is showing us about how it actually runs.
Workflow intelligence: what individual employees are learning as they work, especially with AI.
Pilot intelligence: what active experiments are revealing.
How signal gets processed
There are also four stages of signal processing:
Capture. The signal is recorded.
Cluster. Similar signals form a pattern, a theme.
Conclude. Leadership names the implication.
Codify. The learning is written back into the operating context so it compounds.
Each stage matters, but the fourth one is where most organizations stop. (I know this firsthand, even in my own small organization. This is the hard part.) Signals get surfaced. Themes get discussed. Conclusions get made. Then nothing changes structurally, and next quarter the same signals appear again. The loop only closes when the learning is codified into the playbook, the roadmap, or the strategy itself, in a way the rest of the system can read going forward.
An Example of Closed-Loop
Imagine a sales rep at an industrial rental company. She delivers what she believes is the value proposition: lowest price in the market, ready when the customer needs it. During the call, the conversation turns to the customer’s actual concern, which is access control. The job site is busy, with people moving around the equipment all day, and the customer wants assurance that only authorized operators can fire it up. Price factors in, but price is not what the customer is buying.
In an open-loop organization, that conversation never goes anywhere. The rep might adjust her pitch the next time she gives it, or mention to marketing that some collateral about access control could help drive more leads, but the lesson never compounds the way it should.
In a closed-loop organization, the call is captured and added to the queryable layer. The system clusters it with other recent conversations and surfaces the theme: security is showing up across multiple deals in this segment. Leadership concludes that the value proposition is mis-calibrated, with downstream consequences for marketing, sales, and product. The team codifies the change deliberately.
Maybe the marketing collateral is updated, or sales gets a new discovery sequence, or product reprioritizes the roadmap. Whatever the codification looks like, the next time a rep walks into a similar conversation, what the organization has learned is now with her.
What Closing the Loop Requires
None of this is easy for incumbents. It is a significant undertaking, with legal and privacy risks that have to be taken seriously. What goes into the queryable layer will be different for every organization, but it will likely involve giving AI structured access to call transcripts, meeting notes, internal documentation, shared communication tools, and shared drives.
Just as important, the labor has to be divided correctly. AI captures and clusters. Leadership concludes. The CEO codifies, which is the structural work nobody else has the authority to do. A chief AI officer cannot rewrite positioning. A head of customer success cannot reprioritize product. I have argued elsewhere that AI transformation cannot be delegated, and the closed-loop work is what that argument looks like in practice.
The Real Moat
The companies that figure this out will win because they have built the discipline to turn observation into codified learning, and that learning will compound while their competitors keep relearning the same lessons.
This is the work: capturing the signal, clustering it into a pattern, drawing the conclusion, codifying the learning, and having the patience to run the sequence on a cadence until it becomes how the company thinks. The closed-loop organization is a new way of operating that AI now makes possible.
Whoever builds it first in your market will be very hard to catch.
Related Insights
