DealSeal AI

From conversation to clarity — and from agreement to PayPal checkout

Comment

Inspiration

Every day, freelancers, small businesses, and clients make agreements through WhatsApp, Slack, email, and text messages. A project might start with a simple conversation: “I'll build your website for $2,300 and deliver it by November 20.” But what happens when the scope changes, payment expectations differ, or one party remembers the agreement differently? Traditional contracts can feel too formal and time-consuming for smaller projects, while informal conversations leave room for misunderstandings. We built DealSeal AI to bridge the gap between a conversation and a trustworthy agreement. Our inspiration was simple: What if AI could help two people turn their everyday conversations into clear, reviewable agreements, identify what's missing, and establish mutual approval before payment?

What it does

DealSeal AI is an AI-powered agreement assistant that transforms informal conversations into structured, reviewable agreements. Users provide a conversation between two parties, and DealSeal AI analyzes it to identify important terms such as pricing, delivery deadlines, scope of work, revisions, payment schedules, acceptance criteria, and intellectual property ownership. Instead of assuming everything has been agreed upon, DealSeal AI distinguishes between confirmed, proposed, conflicting, and missing terms. It highlights unresolved questions so both parties know exactly what needs clarification. Once the terms are resolved, DealSeal AI supports generating an agreement draft, managing revisions, recording explicit approvals from both parties, and creating a finalized PDF with integrity verification. We also integrated PayPal Sandbox to demonstrate how payments can be connected to a structured agreement workflow. The goal is not simply to generate a contract. It's to help both parties understand what they're agreeing to before money changes hands.

How we built it

We built DealSeal AI as a full-stack application using modern web technologies, AI, and cloud infrastructure.

  • Frontend: React and Vite for an interactive user experience.
  • Backend: Java 21 and Spring Boot for REST APIs, agreement workflows, validation, and business logic.
  • AI: OpenAI integration to analyze conversations, extract agreement terms, and identify ambiguity.
  • Database: SQLite to persist agreements, versions, approvals, and payment-related demo records.
  • Document generation: Apache PDFBox to generate finalized agreement documents.
  • Payment integration: PayPal Sandbox for demonstrating payment flows associated with agreements.
  • Cloud infrastructure: AWS EC2 for hosting, with S3 and AWS Systems Manager supporting deployment and operations.
  • Security and integrity: Version-specific approvals, SHA-256 hashing, and checks that prevent unresolved terms from silently becoming finalized agreements. One of our key design decisions was to keep AI analysis separate from agreement approval. AI can interpret and organize the conversation, but it cannot provide consent on behalf of either party.

Challenges we ran into

  1. Understanding informal conversations Real conversations are messy. People change their minds, contradict earlier statements, use vague language, or leave important details unstated. Extracting agreement terms reliably required more than simply asking an AI model to summarize a conversation.
  2. Preventing AI from inventing agreements One of our biggest challenges was ensuring that DealSeal AI never treats an AI-generated assumption as confirmed consent. We introduced structured statuses, evidence references, normalization, and validation to distinguish what was actually discussed from what still needs agreement.
  3. Managing revisions and approvals An agreement isn't static. If the price or delivery date changes after someone approves it, the earlier approval should no longer count. We implemented versioned agreements and approval resets to address this.
  4. Connecting agreements with payments Integrating PayPal required us to distinguish agreement approval from payment authorization. We designed the Sandbox workflow around that separation rather than assuming that accepting an agreement automatically authorizes a payment.
  5. Building and deploying the complete experience Connecting the React frontend, Spring Boot backend, AI integration, persistent storage, PDF generation, and AWS infrastructure into a functioning demo introduced challenges around configuration, API integration, testing, and deployment.

Accomplishments that we're proud of

Our biggest accomplishment is building more than an AI text summarizer. We created an end-to-end prototype that connects conversation analysis with a structured agreement workflow.

We're particularly proud of:

  • Evidence-based AI analysis: Agreement terms are tied to conversation evidence rather than unsupported assumptions.
  • Missing-term detection: DealSeal AI actively identifies gaps instead of producing a polished-looking but incomplete agreement.
  • Human-controlled approval: Both parties must explicitly approve the relevant agreement version.
  • Version integrity: Changes to an agreement invalidate earlier approvals.
  • Verifiable documents: Finalized PDFs can be checked against their recorded hashes.
  • PayPal integration: We demonstrated how agreement clarity and payment workflows can work together.
  • Cloud deployment: We built and hosted a working full-stack prototype rather than stopping at a local proof of concept. Most importantly, we designed the product around a principle we believe matters: AI should help people reach an agreement, not decide that they've agreed.

What we learned

Building DealSeal AI taught us that applying AI to real-world workflows requires much more than generating accurate text. We learned that trust, traceability, and explicit human decisions are just as important as model intelligence. We also discovered how challenging it can be to convert unstructured conversations into structured information without losing context or overstating certainty. From an engineering perspective, we gained experience combining AI-generated output with deterministic validation, implementing version-aware workflows, integrating PayPal Sandbox, generating verifiable documents, and deploying a full-stack Java and React application on AWS. Perhaps our most important lesson was that the best use of AI isn't always automation. Sometimes its greatest value is making ambiguity visible so humans can make better decisions.

What's next for DealSeal AI

We envision DealSeal AI becoming a lightweight agreement and payment platform for freelancers, creators, consultants, agencies, and small businesses. Our next steps include:

  1. Real-time negotiation assistance: Help both parties resolve missing or conflicting terms through guided questions.
  2. Conversation integrations: Support importing discussions from email, Slack, and other messaging platforms.
  3. Reusable agreement templates: Offer workflows tailored to website development, design, consulting, and other freelance services.
  4. Verified identities and electronic signatures: Move beyond demo approvals toward stronger identity verification and legally appropriate signing workflows.
  5. Deeper PayPal integration: Explore payment milestones, invoices, and payment-status tracking tied to agreement terms.
  6. Production-grade infrastructure: Strengthen authentication, access controls, audit trails, and scalable persistence. Our long-term vision is to make creating a clear agreement as easy as having a conversation.

DealSeal AI — From conversation to clarity. From clarity to agreement. From agreement to confident payment.

Built With

Share this project:

Updates

Submission history