Vasuman “Vas” Moza is the founder and chief executive of Varick Agents, which builds enterprise AI agents around the messy realities of how companies operate. His central proposition: increasingly capable models can execute tasks, but reliable automation depends on understanding informal handoffs, operational exceptions, existing software, and which decisions require human judgment.
Moza cofounded Fremont Debate Academy in 2014 and served as its chief technology officer. He studied computer science and business at the University of California, Berkeley, appeared in its 2022 commencement program, and spent three years as a software engineer at Meta before leaving to found Varick Agents.
Building enterprise agents around actual work
- Business context is the bottleneck. Moza considers customer-specific organizational knowledge more difficult to acquire than the ability to complete isolated tasks. Official documentation routinely misses the dependencies, interpersonal handoffs, and exceptions that determine whether an automated workflow survives production.
- Map the workflow before automating it. Varick embeds engineers and strategists with customer teams to examine processes including accounts payable, accounts receivable, billing, reconciliation, banking, and financial planning. They investigate routine operations and failure cases, then divide responsibilities among autonomous agents, human reviewers, and people retaining direct control of consequential decisions.
- Keep existing systems of record. Through Varick OS, agents operate above established platforms including NetSuite, Microsoft Dynamics, SAP, and Salesforce. Companies can build and monitor automated workflows without abandoning the infrastructure their operations already depend on.
- Scale forward-deployed engineering with AI. Moza wants internal assistants to help customer-facing engineers absorb documentation, track changing requirements, and maintain organizational context across engagements. The objective is deeper customer support without increasing specialist headcount at the same rate.
In his public guide to forward-deployed engineering, Moza emphasizes technical execution, business diagnosis, communication with nontechnical stakeholders, and narrowly scoped initial deployments. His AI Engineer presentation places those ideas within a practical operating sequence: audit how a business works, redesign selected workflows, and expand agent autonomy as operational trust develops.
