Sirius, a San Francisco AI startup, has reached $3 million in annual recurring revenue with a platform that intercepts subscription cancellations and delivers personalized retention offers in under 200 milliseconds, according to a company disclosure published August 21, 2026. The firm, which emerged from Andreessen Horowitz's a16z Speedrun Cohort 006, operates with just 15 employees and positions itself as a pioneer in what it calls autonomous retention software. By shifting from after-the-fact analysis to split-second intervention during the cancellation flow itself, Sirius is targeting a critical profit driver for subscription businesses.
The company's cancellation decision engine processes millions of customer interactions each day, evaluating intent and returning recommendations fast enough to insert offers before a subscriber finalizes their exit. With $3 million in ARR spread across a 15-person team, Sirius is generating roughly $200,000 in annual recurring revenue per employee. Early customers include Higgsfield, Headway, and Promova, all of which also participated in the a16z Speedrun Cohort 006 alongside Sirius.
The report states that Sirius is attempting to create an entirely new category it describes as autonomous retention, one that merges analytics with execution rather than treating them as separate steps. According to the disclosure, the platform embeds its decision engine directly into the cancellation workflow, transforming what was traditionally a loss event into a potential retention opportunity in real time. The report notes that this approach mirrors the wider shift toward product-led growth, where the product itself becomes the main channel for both sales and keeping customers on board.
Sirius enters a market already occupied by established players such as Gainsight, Totango, and Stripe Billing, which control data pipelines and offer post-cancellation incentives but lack the ultra-low-latency decision layer the startup highlights. The report argues that if Sirius can show its interventions boost net retention without eroding revenue through unnecessary discounts, it may compel incumbents to either partner with or acquire similar technology. While the $3 million ARR figure remains modest, the disclosure suggests it demonstrates unit economics capable of rapid scaling once integration friction decreases. The report warns that the central challenge ahead will be whether Sirius can preserve the accuracy of its AI models as it expands into sectors with different pricing models and customer lifecycles. Success, the report concludes, would validate a new software category and reinforce the idea that AI-native, micro-decision tools represent the next frontier for subscription efficiency, with investors and operators advised to watch the company's product roadmap and its capacity to secure long-term contracts that make the decision engine a core piece of the subscription infrastructure. For companies weighing whether to build or buy retention intelligence, the choice now includes a third path: embedding a specialist layer that operates faster than human teams can react. That trade-off between control and speed will likely define competitive positioning in subscription markets far beyond SaaS.