To encourage employees to automate tasks, the process of creating the automation must be demonstrably easier and faster than performing the task manually. Otherwise, people will always default to the path of least resistance, which is the manual action.
For enterprises, the raw capability of foundation models is a security risk, not a selling point. The real product value lies in building "boundaries"—robust permissions, approvals, and audit logs that make powerful models safe to deploy company-wide.
Making automation too easy can lead to "slop"—numerous duplicate and poorly managed workflows. Serval solves this with an AI agent that understands existing automations, preventing redundancy and suggesting consolidation or modification instead of creating new, duplicative workflows.
Individual employees want powerful, autonomous AI agents similar to consumer products. However, the enterprise prioritizes control, safety, and governance. This creates a fundamental tension that enterprise AI products must navigate, balancing user desire for freedom with the organization's need for security and oversight.
With foundation models making technical features easy to copy, the sustainable advantage for AI companies lies in deep customer understanding. Serval's CEO stays in over 100 customer Slack channels daily to build this "customer insight" moat, which is harder to replicate than any product feature.
A multi-model strategy is key. Serval finds that OpenAI's models consistently excel at user-facing interactions and correctly calling tools. For backend code generation to create automations, however, Anthropic's models currently deliver superior performance, highlighting the need to match models to specific applications.
Building an AI-native organization means questioning the need for entire departments. Serval starts with the assumption that a role could be handled by AI, giving it the "right of first refusal." This has allowed them to eliminate traditional roles like Solutions Engineers and SDRs, empowering AEs with AI tools instead.
Unlike companies that resell tokens for every query, Serval uses expensive models once to create a durable script. This automation is executed repeatedly at low cost. This "generate-once, run-many" approach dramatically improves unit economics and insulates the business from high token consumption.
Integrating the latest foundation model is complex because new models can break prompt tuning built around the quirks of older versions. Serval has found that a new model's unpredictability can outweigh its intelligence, sometimes forcing them to downgrade to an older, more reliable model to ensure consistent behavior.
To balance power and safety, Serval uses two distinct agents. An "Admin Agent" helps IT build and approve workflows with specific permissions. A separate "Help Desk Agent" for end-users can only execute these pre-vetted tools, allowing it to "run wild" within a secure, pre-defined sandbox.