Deploy AI agents in minutes — no developers required

As a technology partner in design and software engineering since 2020, we have been responsible for the evolution of a U.S. cyber risk platform—from prototype to a mature SaaS product used by Fortune 500 organizations.

In 2025, we rebuilt the runtime environment (runtime), basing it on AI agents powered by Anthropic’s Claude models. Today, the platform serves as the orchestration layer for the client’s entire cybersecurity stack.

From prototype to an Enterprise SaaS platform

The client—a cybersecurity company founded in 2015—approached us with a prototype of a decision engine for the executive team (C-level), complete technical documentation, and its first paying enterprise customers. However, the software operated solely in an offline (on-premise) model. Our goal was to turn it into a scalable SaaS application without slowing the client’s market momentum.

Product discovery workshops helped surface and prioritize key features. We designed and implemented a modern frontend that translates complex mathematical risk models into clear, immediately actionable business recommendations (actionable insights), aligned with NIST and CIS standards. Given the stringent requirements around telemetry data confidentiality, we configured an isolated GCP cluster in a dedicated location specified by the client.

After a successful MVP launch, we expanded the target audience to include private equity funds. In just 3.5 months from launch, the enterprise customer base more than tripled (from 8 to 26 organizations), encompassing industry leaders and Fortune 500 corporations. Revenue grew 150% YoY, the B2B customer base increased by 400%, and NPS consistently stayed above strategic targets.

The 2025 challenge

Despite a strong market position, the product encountered barriers:

• High barrier to entry: Reports required advanced expert interpretation.
• Low daily retention: Most users logged into the system only once per quarter.
• Time-consuming onboarding: Onboarding a new user took 45 minutes and required a consultant’s assistance.

The client decided to implement agentic artificial intelligence (Agentic AI). Due to the stringent security requirements of enterprise customers, a simple chat window did not meet expectations. The new AI module had to:

• Operate reliably in production within a secure multi-tenant architecture.
• Process telemetry and event data streams in real time.
• Remain neutral to the existing certification and compliance audit scope.

Architecture: Orchestration and Security by Design

We designed the runtime environment around three key pillars of Enterprise AI deployments: the agent’s reasoning mechanism, isolation at the security perimeter, and direct integration with the client’s technology stack.
 

Solution: Overarching orchestration layer

The core infrastructure, data model, and platform codebase remained intact – the AI layer was deployed as a modular orchestration overlay. As a result, the product transformed from a static analytics tool into a security operations orchestration (SecOps) platform.

Expanding the platform’s capabilities has been decoupled from traditional release cycles: the management team can roll out new prompts, connectors, and specialized agents directly via the GUI in the production environment.