# AI tools for Forward Deployed Engineering — Vasuman Moza, Varick Agents

## Executive summary

The core bottleneck in enterprise AI adoption is not model intelligence or execution capability, but rather the ability to deeply understand and re-engineer complex, undocumented human business processes. Forward Deployed Engineers (FDEs) are critical for this process, mapping existing workflows—including edge cases and handoffs—and building autonomous agents on top of legacy systems (e.g., SAP, NetSuite). The technical solution involves using a Veric OS platform that builds agents on existing systems without requiring costly migrations, leveraging dependency graphs, and employing custom post-trained models to extract accurate context from messy enterprise data.

## Key takeaways

- The AI Bottleneck is Context, Not Execution: While modern LLMs can solve the execution of work (intelligence constraint), the primary bottleneck remains understanding the unique business processes within a specific company. Every department (e.g., healthcare sales vs. SaaS sales) operates differently, requiring deep context extraction.
- Forward Deployed Engineering (FDE) Role: FDEs are responsible for mapping how humans currently perform work and then re-engineering the process around AI. This ensures that AI solutions are adoptable and deliver measurable, department-wide ROI rather than failing as isolated 'point solutions.'
- Non-Disruptive Deployment Strategy: To overcome enterprise resistance to migration (e.g., spending $5 million over five years on NetSuite), agents must be built *on top* of existing systems of record (Salesforce, SAP, Dynamics) rather than requiring a full system overhaul.
- Agent Tooling for FDEs: The Veric platform provides specialized tools: an Engagement Agent (assistant to synthesize notes/docs), and a Workflow Agent that ensures the constructed workflow correctly shadows real-world process edge cases. A future autonomous assistant will handle minor changes without human intervention.

## Technical details

- Workflow Modeling: Workflows are modeled using a dependency graph, which is preferred over traditional database structures because enterprise processes are often cyclical and require clear dependencies (e.g., Person C cannot proceed until A and B approve).
- Custom Model Training: To improve output quality, the system uses post-trained models (on open-source bases like Kimmy K26) to achieve a balance between detail and clarity that general frontier models often lack.
- Context Extraction & Graph Traversal: The most complex technical challenge is reliably traversing the knowledge graph. This requires training custom tools within an RL environment to solve problems like entity resolution (determining if 'Person A' and 'Person B' are the same person) and identifying redundancy cycles.
- Platform Architecture: The Veric OS platform allows agents to be spun up, monitored, and governed while remaining layered on top of existing enterprise systems (e.g., SAP, Salesforce).

## Practical implications

- AI implementation must prioritize process re-engineering and human workflow mapping over simply applying AI to broken processes.
- Architectural solutions should favor overlay agents on existing systems (System of Record) rather than mandating costly migrations.
- The complexity lies in context management: building tools that reliably traverse knowledge graphs and resolve entity ambiguity is critical for autonomous agent function.

## Topics

AI Agents, Forward Deployed Engineering, Enterprise Architecture, Process Automation, Knowledge Graphs, Veric Agents, Claude / Codex

Source: https://www.youtube.com/watch?v=l0FLhNqBOic