ScopeLock

Inspiration

I got this inspiration while I was doing AI and Data consulting/freelancing work. Proposal creation was never the hard part since I already had an SOP and everything needed to guide the proposal. But ongoing communication with the client sometimes makes the project scope keep expanding without me realizing how far we have come from the initial agreement. This is what slowly kills the margin. So ScopeLock was introduced to help freelancers and service providers protect their scope and monetize their skills more effectively.

What it does

ScopeLock watches your email inbox for incoming client requests. It autonomously analyzes the intent and requirements of the email, checks them against your SOP, and generates a proposal automatically. The work does not stop there. ScopeLock continuously monitors the email thread, tracks changes to the project scope, calculates the corresponding cost impact, and prepares an updated proposal whenever the scope changes.

How I built it

I built ScopeLock using Gemini 3.5 Flash and Google Agent Development Kit (ADK) for the agent workflow. Gmail events are delivered through Google Cloud Pub/Sub to a backend running on Cloud Run, while Firestore stores project requirements, scope history, agent runs, and proposal states. The frontend is built with Vite, React, TypeScript, and Tailwind CSS. ScopeLock prepares proposals autonomously, but users must explicitly review and approve them before they are sent back to the client.

Challenges I ran into

The biggest challenge was making sure the AI could understand whether a client request was actually new scope or just a clarification of something already agreed on. I also had to separate AI reasoning from pricing calculations so Gemini understands the request, while deterministic rules based on the SOP handle the actual price and timeline.

Accomplishments that I'm proud of

I'm proud that ScopeLock is not just another chatbot or proposal generator. It can react to real client emails, perform a complete workflow autonomously, remember the agreed scope, and continue monitoring the project as the conversation evolves.

What I learned

I learned that building a reliable AI agent is not about letting AI make every decision. AI is great at understanding messy human communication, but important business decisions like pricing, scope, and approvals still need structured rules and clear boundaries.

What's next for ScopeLock

Next, I want users to be able to upload and customize their own SOPs, pricing rules, and service packages. I also want to expand ScopeLock beyond Gmail into tools like Slack, CRM systems, and project management platforms so it can protect project scope wherever client communication happens.

Built With

• fastapi
• firestore
• gcp
• gemini
• gmail
• google-cloud
• google-gmail-oauth
• python
• react
• typescript
• vertex
• vite