Matt Slotnick: the next enterprise giants will own intent, not records

"Intention Is All You Need", an essay by Matt Slotnick, co-founder and CEO of enterprise AI startup Poggio Labs, argues that the software category that has minted the largest enterprise companies of the past thirty years — the system of record — is about to be demoted to an input. Kasey Klimes, who builds decision infrastructure at Primitive, called it "easily one of the best writings on AI this year," which is how it resurfaced six months after publication.

Slotnick's starting premise is conventional: businesses are resource allocation machines, and the most valuable software owns the objects those resources are allocated against. "Salesforce owns the customer. Workday owns the employee. ServiceNow owns the technology." Those objects are not merely databases, he writes, but "representations of organizational truth" — definitions of what a customer is, how it is measured, which states it can occupy. Thousands of people coordinate because they share the same map of Leads, Opportunities, Accounts, Tickets, which is also why seats became the unit of pricing.

#From objects to objectives

The break comes with a new resource to allocate. Agents, in Slotnick's framing, are "infinitely scalable, programmable, and if you believe the marketing, will soon be infinitely capable." Once they are in the mix, the design question changes from "how do we give people a shared map to decipher?" to "how do we give agents and people a shared destination to work towards?"

Shared understanding of objects becomes necessary but no longer sufficient: "What the customer is matters less than what needs to happen with the customer. The current system describes state." What wins instead is a system that defines and manages intent — the outcome the organization is pursuing plus the plans to get there. Close this customer. Resolve this issue. Hire for this role. Systems of record created gravity by owning shared understanding; the next generation, he argues, will create it by owning shared purpose, and that pull will be stronger because such a system does not merely describe workflows, it executes them end to end.

The intention-centric architecture: context at the bottom, intention in the middle, action on top. Credit: Matt Slotnick, "Intention Is All You Need".

#The map is not the territory

The sharpest section is an argument about compression. A Lead record flattens a human being into a handful of fields; an Opportunity flattens months of negotiation into a stage and a dollar amount. That compression existed because people could not process the full picture. Agents can: call transcripts, email and Slack threads, meeting notes — the unstructured exhaust that piled up for years because it was too dense to synthesize. "The map was a required compromise. Agents can navigate the territory."

Intent, in his architecture, is not a goal on a slide but a structured object: a specific outcome, its constraints, its assumptions, and the evaluations that judge it. "A goal you can't decompose into tasks, measure against outcomes, or adjust based on new information isn't intent, it's a pipe dream." His example of the difference specificity makes: "close this customer" versus "close this customer by demonstrating ROI in their compliance workflow, leveraging the CFO relationship, before FY budget cycle."

#The bottleneck inverts

The economic claim behind all this is that execution stops being the expensive part, and the bottleneck inverts from doing the work to defining it — "agent execution scales infinitely, but human task generation does not." The system that wins, on his account, closes its own loop: observe state, define the objective, plan, act, evaluate, adjust. More context yields better tasks, better tasks better outcomes, better outcomes more context. He is not romantic about the risk: at unbounded speed, small errors compound and "send you disastrously in the wrong direction."

Software development is his proof point, and his explanation for why it went first is the essay's tidiest idea: code is the one domain where the territory is already digital and legible, so no compression into objects was ever required. He cites OpenAI's harness engineering post — a small team steering Codex to 1,500 merged pull requests with zero manual coding — and Spotify's co-CEO saying senior engineers there have largely stopped writing code themselves. "Other domains must discover the abstractions, and create legibility. The platforms that do so will win."

#Why it matters

Two caveats belong on this. Slotnick is selling the thesis: Poggio builds account intelligence for revenue teams, so "the intent layer subordinates the system of record" is also a pitch deck. And the essay is a design argument, not evidence — outside coding, no one has shown that an intention layer can hold a moat against the incumbents whose data it depends on. What makes it worth reading anyway is that it names the thing most agent products are quietly groping for. Everyone is building context pipelines; far fewer are building a durable, structured representation of what the organization is trying to do — the artifact an agent needs before autonomy is anything but expensive improvisation.