10X reports 50% Fortune 2000 client base and 2x output growth

10X, an applied AI partner based in New York City, has grown to 75 employees, 90% of whom are engineers or AI strategists. The company reports that 50% of its customers are now Fortune 2000 companies, while the other 50% are mid-market firms with $100 million to $500 million in revenue. Clients increasingly request 'AI transformation as a service' rather than simple staff augmentation. 10X engineers report a 2x increase in output over the past year, driven by the adoption of models like Opus 5.5 and tools like AMP, which reduced costs by 80% via Cloud Code subscription integration. The company utilizes a 'meta-harness' consisting of standardized documentation, a CLI, and Git-checked-in logs to drive agent workflows. This system allows agents to log decisions and maintain durable records of context, replacing reliance on individual memory or ephemeral conversations. CJ, a special projects lead at 10X, notes that models can now reach 90% of a solution given the right context, shifting engineer value toward customer interaction and architectural oversight. 10X identifies a significant 'diffusion gap' in enterprise AI adoption. While 90% of the 100 enterprises they consult with have rolled out general-purpose LLMs like Claude or Codex to employees, and 60% have implemented training and enablement, only 50% have deployed coding agents to engineering teams. The most significant drop-off occurs in non-engineering work, where only 5% of large enterprises have agents in production for tasks such as back-office operations. Key barriers include permissioning and RBAC challenges, which create security nightmares for CIOs and CISOs, and the lack of standardized processes. 10X addresses these issues by building bespoke gateway solutions for data permissioning and helping clients standardize workflows before deploying agents. 10X argues that the 'final mile' of AI diffusion is the hardest problem, requiring specialized partners to help enterprises move from AI-absent to AI-native, and that output-based compensation models are the future of engineering.