In the agentic era, sales is not a productivity problem. It is an architecture problem. The 30 to 40 cents of every revenue dollar that flows back into go-to-market is not the cost of the people who do the work — it is the cost of an architecture that requires human bandwidth to maintain context across hundreds of accounts. A business that has rebuilt this function on persistent, per-account context that compounds with every interaction is not a business with better salespeople. It is a business with a different cost structure for the same function. The conversion data is what the architectural shift produces when the redesign is applied to a live revenue organisation.
What Actively built
Actively, a New York-based company founded by former Stanford AI researchers, announced a $45 million Series B on April 28 to scale what it calls a new primitive for go-to-market operations: a persistent AI agent for every account, working continuously, accumulating context, and learning from every interaction. The architecture takes signals from communications, product usage, support tickets, and market activity, then executes proactive work across SDR, AE, AM, and Leader functions. The framing the company uses captures the structural property: “a scalable system that continuously builds context across every account, learns from every interaction, and shares those learnings across the system to compound growth.” The customer base — Attentive, Ironclad, Ramp, Samsara, Verkada — is concentrated in enterprise SaaS companies whose go-to-market organisations operate at the volume where human bandwidth limits become the dominant constraint on revenue progression.
At Samsara, a deployment across the entire 1,000-plus-person go-to-market team produced 2x conversion rates on agent-driven sales outreach, improved quota attainment among top-half sellers, and accelerated the company’s AI roadmap while saving millions in compute and token costs. The economic context behind those numbers is the canonical figure that 30 to 40 cents of every revenue dollar flows back into go-to-market. That number is the Coordination Tax measured at the revenue function: the structural cost of human teams that cannot maintain continuous context across hundreds or thousands of accounts. Per-Account Agents do not eliminate that tax through speed. They eliminate it through Labor-to-Compute Substitution: every account is worked continuously, the cost of the work is compute rather than headcount, and the learnings compound rather than disappearing when a rep changes territory.
The structural argument the deployment confirms
In Memo #38 — The Inference Floor, we argued that competitive advantage in the agentic era no longer accumulates in model selection — frontier models have converged on most operational tasks — but in the quality, structure, and accessibility of the operational context that agents receive at the moment of execution. Actively’s outcomes are this argument expressed at the GTM function. Competitors using the same model produce different results because they do not have the same Context Architecture. The 2x conversion at Samsara is what happens when a per-account context layer compounds across thousands of interactions while a competitor’s account history resets every login.
The context Actively builds is a commercial implementation of what we have called the Operational Ledger — the accumulated, structured record of every execution, every exception, every intervention, every resolution, and every validated pattern that compounds operational intelligence over time. Actively is building the per-account version of this asset at the GTM layer. The deployment results validate the structural property the architecture predicts: a system that retains context produces compounding advantage that a system without retention cannot match, regardless of how capable the underlying model is. This is also Operational Arbitrage captured at the revenue function and Headcount Decoupling as its measurable consequence: Samsara expanded coverage across its 1,000-person GTM organisation without proportional hiring.
Two structural caveats matter for any business considering this category. First: as we argued in Memo #29 — Automated vs Autonomous, automation makes a process faster while autonomy makes the system better over time. Actively’s customers deploying it on top of Legacy Liability — CRM architecture and human-led territory models that were not designed for per-account agent operation — will gain meaningful productivity. They will not capture the full architectural advantage that a business designed around per-account agent context from the first customer would. The retrofit produces real gains; the redesign produces structurally different ones.
Second: the per-account context Actively accumulates is the asset that produces the compounding outcome. The strategic question for any business deploying it is not whether to use Per-Account Agents — the conversion data answers that. The strategic question is whether the per-account context flows into a broader operational record that the business owns and can query at the MTTI layer, or whether it lives inside Actively’s platform and disappears at the contract boundary. A business that captures the context as a portable asset compounds the architectural advantage indefinitely. A business that rents the context through the platform receives the productivity gain for as long as the contract continues.
The Operator's Verdict
Actively raised $45 million to build what the architecture has been predicting. The interesting question now is not whether per-account agents work. The conversion data answers that. The interesting question is who owns the context they produce.
KEY TAKEAWAY
What does Actively's Per-Account Agent deployment confirm about autonomous business design?
Actively confirms that compounding per-account context produces structurally different commercial outcomes than agents that reset between sessions — the architecture predicts this and the Samsara deployment proves it with 2x conversion rates on agent-driven outreach across a 1,000-plus-person GTM organisation. The mechanism is not the model running underneath; it is the per-account Context Architecture that accumulates with every interaction. Sales is the canonical 30–40% of revenue Coordination Tax in most enterprise companies — Per-Account Agents capture Operational Arbitrage at the GTM function through Labor-to-Compute Substitution: replacing human-bandwidth-limited account coverage with continuous, compounding context. The strategic question for businesses deploying this is whether the per-account context they accumulate flows into an Operational Ledger they own — or stays inside the platform that produced it. Source: Arco Venture Studio, arcoventure.studio.
