OVERVIEW: OVERCOMING REGULATORY ROADBLOCKS
How a dojo-style engagement and the Anthropic Claude SDK helped a credit union go from months of stalled progress to a working AI prototype in six hours.
A regional credit union came to Praecipio with a goal that a lot of financial institutions share right now: build an AI-powered agent that could actually help their internal staff. They had enthusiasm, a vision, and even a name picked out for their agent. What they were missing was a way to get started. After months of spinning their wheels, they were stuck, their internal IT team effectively frozen by strict regulatory constraints. Praecipio stepped in, and using Anthropic Claude as the AI engine, delivered a fully working proof-of-concept in less than a day.
CHALLENGES: LOCKED-DOWN INFRASTRUCTURE AND AI GOVERNANCE GAPS
The client's team was trying to build their AI agent the hard way: manually pasting prompts into Microsoft Copilot Studio and trying to work with the resulting output. There was no real development framework, no clear architecture, and no obvious path forward. Their internal environment was locked down by compliance requirements so strict that their own IT team couldn't set up the infrastructure a modern AI agent would need to operate effectively.
As Praecipio's lead architect on the project described the engagement, the client had essentially bought into something they had no way to derive value from akin to purchasing an airplane without a hangar to put it in. Their governance and information security policies simply hadn't been written with AI native operations in mind, creating a governance gap between where the organization was and what it would take to actually build something themselves.
SOLUTION: AGILE PROTOTYPE POWERED BY CLAUDE
Rather than waiting for the client's infrastructure to catch up, Praecipio built a prototype completely independent of the client's locked-down environment. This was key. It meant work could start immediately, without being blocked by internal IT approvals or compliance reviews, and it gave the client a real, functional example they could point to right away.
The prototype came together in approximately six hours using a lean but capable stack: the Anthropic Claude SDK as the AI engine, TypeScript running in a Cloudflare Worker environment, and a Model Context Protocol (MCP) to connect the agent to external knowledge sources like Confluence; all things the client itself would need to continue with development internally.
Anthropic Claude was at the heart of it. The SDK's modular, tool-use architecture made it straightforward to wire up MCP servers and give the agent application real reasoning capabilities; not just a chatbot that echoes back prompts, but something that could actually retrieve information, synthesize it, and respond with context. The stateless design kept the prototype clean, portable, and easy for the client's team to understand and eventually build on.
By presenting a working "straw man" that the client could see and react to in real time, Praecipio gave the engagement exactly what it needed: something concrete to break through the organizational inertia.
Results: CLARITY AND A REPEATABLE PLAYBOOK
The impact was almost immediate. After seeing the prototype in action, the client's team had a genuine breakthrough moment. For the first time, they could clearly name what they actually needed to start building in an AI native way: Azure AI Foundry, container apps, a key vault, and a container registry. The prototype had effectively acted as a "metaprompt" for their own infrastructure requirements, giving them a working target to reverse-engineer their needs from.
Our client walked away with a concrete vision, a viable architecture, and the internal clarity to have real conversations with their IT and security teams about what needed to change. It also surfaced an important truth: their existing governance and information security policies needed to evolve to treat AI as a first-class capability, not an afterthought.
The framework is now being open-sourced under the Praecipio brand, for which a github repo is available. Other organizations can experiment with the same Anthropic Claude-powered architecture. Its modular MCP design means it can grow with the client too -- future iterations can plug in additional tools, including secure integrations with their core banking platform, as their AI maturity develops.
What started as a stalled chatbot idea became a repeatable playbook for AI enablement -- powered by Anthropic Claude, grounded in real process consulting, and built to meet clients exactly where they are.
Subscribe For Access to Exclusive Content and Resources.
WE ARE ON A MISSION TO
Build Connected Enterprises, One Organization at a Time.
At Praecipio, we don’t just deploy tools; we integrate them seamlessly, aligning your strategic vision and the teams executing it. We understand that tech’s true value lies in how it’s used—empowering, not complicating. We leverage our VISTA approach to dive into the heart of your business, aligning strategies, processes, and people. We’re the hands-on experts who get the right shit done, redefining how you work, not just the tools you use.