Inspiration
Freelancers and clients constantly face a trust deficit. Freelancers worry they won’t get paid after submitting work, and clients worry they’ll pay for deliverables that don't match what was promised. Existing solutions are either closed marketplaces that take massive commission cuts or simple payment gateways that don't verify the actual work. We were inspired to build a neutral transaction assurance layer—where AI checks the work against the initial agreement before the money moves.
What it does
Dilio is an AI-powered transaction assurance platform for independent professionals. It acts as a coordination layer for parties who have already found each other.
- AI Agreement Builder: Users paste their messy chat logs or project briefs, and the AI extracts deliverables, deadlines, and suggests structured payment milestones.
- Milestone Payments: The buyer funds individual milestones via a PayPal integration.
- AI Delivery Verifier: When the freelancer submits work, the AI checks the evidence strictly against the agreed-upon acceptance criteria. It provides a verification report (Met/Not Met) and recommends acceptance or revision.
- Accountable Payouts: Only after the buyer reviews the AI report and manually approves the work is the PayPal payment released.
How we built it
We built the frontend wireframe and UI architecture using React, TypeScript, Tailwind CSS, and shadcn/ui to ensure a highly professional, trustworthy, fintech-oriented design. The platform is structured around a dual-status system: tracking the Dilio workflow status (e.g., "Work Submitted") completely separately from the PayPal payment status (e.g., "Captured" or "Released"). For the AI components, we designed specific prompts to handle term extraction and evidence-based verification without inventing missing information.
Challenges we ran into
Our biggest challenge was balancing AI capabilities with human control. We initially wanted the AI to automate the entire process, but realized that financial transactions require human confirmation. We had to strictly constrain the AI so it only acts as an advisor (recommending acceptance or revision) rather than an executioner that moves money. Designing an interface that cleanly separates the project workflow from the underlying financial ledger without confusing the user was also a major UI/UX hurdle.
Accomplishments that we're proud of
We are incredibly proud of the AI Delivery Verifier. Watching the system successfully map a submitted piece of evidence back to a specific line-item acceptance criterion in the contract feels like magic. We're also proud of the clean, PayPal-inspired visual identity that makes the platform feel like a serious, production-ready SaaS product rather than a basic hackathon prototype.
What we learned
We learned a lot about prompt engineering for commercial contracts—specifically, how to stop large language models from hallucinating commercial terms when information is missing from a project brief. We also deepened our understanding of milestone-based payment architectures and how to handle edge cases like disputes, revisions, and partial funding within a UI.
What's next for Dealio
Our immediate next step is taking the frontend architecture and connecting it to a robust backend using Supabase and the live PayPal API (moving from the Sandbox environment to a production flow). We plan to expand the AI Delivery Verifier to handle more complex evidence types, like scanning GitHub repositories for code deliverables or parsing large Figma files for design criteria. Ultimately, we want to establish the necessary licensing partnerships to offer full production fund custody.
