Work Bench

Discovery // Identifying the core pain point signals

Traditional agencies often suffer from information silos and disorganized workflows, caused by relying on a scattered ecosystem of third-party tools that don’t communicate effectively. Project delays stemmed not from slow development, but from constant context-switching and lost context inherent in standard enterprise tools, which force agencies to conform to the tool.

To overcome this, we leveraged AI capabilities to build our own operations platform, Work Bench, from scratch. Our hypothesis was that a custom build could be deployed as quickly and cost-effectively as integrating a SaaS product. We created an extensible architecture where all modules seamlessly share data and AI context within a single ecosystem built on React, Node.js, and Supabase.

The primary goal of Work Bench was to significantly increase operational efficiency and serve as an internal testing ground for new concepts. As a fully customizable internal application, it is continuously updated to meet evolving needs. To support this rapid pace of development, we established new infrastructure, defining resource allocation and distinguishing between when to use cloud versus local Large Language Models (LLMs) inside the workflow.

In The Weeds // The Mechanics of the GenAI-Augmented Lifecycle

Work Bench is a bespoke, web-based internal tool designed to structure and optimize how an organization operates. Built on a refactored extension of the initial Spark framework (initially created by Peter Taylor-Brown), it was further developed using an accelerated, AI-assisted approach to quickly meet user needs, ensuring every feature has a clear, functional business necessity. Core capabilities are organized into three modules.

1. Product Management: Manages actions for product development and organizational processes, prioritizing full transparency, a nimble product approach, and reduced organizational clutter.

2. Relationship Management: Focuses on client interactions to strengthen relationships, clarify next steps, and surface opportunities, utilizing integrated AI to enrich and correlate data.

3. Experiment Management: Offers a secure sandbox for teams to rapidly iterate on and implement new or existing concepts without facing major technical hurdles.

Previous slideNext slide

Our development process shifted from traditional, slow cycles to rapid, using AI-assisted development tools, like Cursor. We used a 70% AI / 30% human split—AI was responsible for the heavy lifting of code generation, while human expertise directed overall architecture, security protocols, and business logic.

The robust, scalable tech stack includes Vercel (React) for the frontend, Render (Node.js) for the backend, and Supabase for the database, all synchronized via GitHub, designed to grow from supporting a single team to an entire organization.

Business Outcomes // Quantifying the Impact of AI Integration

After just four days of development, we launched Work Bench, a proprietary, AI-powered internal platform. This solution entirely eliminates the need for expensive SaaS applications and was built on existing, scalable, and cost-effective infrastructure, resulting in zero operational expense at launch.

Work Bench is a unified platform that replaces dozens of disconnected tools, allowing teams to centrally manage day-to-day tasks, client conversations, and experiments of ideas in one place. We measure our success by the complete reduction in administrative context-switching. This unification has drastically streamlined our workflow, cutting the time required to move from a raw client inquiry to actionable, prioritized development sprints from weeks down to a matter of days. Crucially, Work Bench is designed to support continuous experimentation, enabling fast responses to emerging user pain points and opportunities.

Takeaways // Lessons from Applying AI in the Wild

• Scalability & Future-Proofing: Work Bench is a scalable framework that future-proofs business operations by allowing for rapid deployment and integration of new, experimental AI-driven tools.
• AI-Empowered Development: AI empowers non-technical users to quickly develop production-grade applications, while custom code ensures complete control and eliminates reliance on third parties.
• Value Proposition & Methodology: The platform's true value lies in instantly translating abstract business logic into functional, enterprise-grade architecture, a methodology applied to all client builds and reinforced by managing projects on their own self-built, AI-first platform.