AI Enterprise SolutionsGPT & Claude

Signflow

Signflow is built for mid-sized companies in fast-moving industries like retail, logistics, agencies, and services that regularly execute contracts without dedicated legal teams. Operations, HR, and finance leaders need to manage agreements safely, while lawyers use it to streamline contract reviews. Finance and operations leads pay for the service because avoiding bad contracts directly protects their budgets. What goes wrong for them today These companies routinely sign long contracts after checking only prices and dates, accidentally agreeing to hidden traps like automatic extensions or uncapped liability. Once signed, contracts sit unmonitored; when laws change or notice deadlines pass, companies face unexpected fines or forced renewals. Their only choices are paying expensive outside lawyers for every agreement or signing blindly and taking on massive financial risk.

Open the demo Video GitHub

Preliminary score

62.0/ 100Promisingaverage of 2 judges

Value for the user18.5 / 25
A specific problem and a clear outcome from solving it.
Prototype and use of AI20.0 / 30
A working core scenario, and what AI actually contributes to the result.
Quality testing9.0 / 20
Tests, examples of failures, and comparison with the current approach.
Feasibility9.0 / 15
Data requirements, running costs and a clear next step.
Originality5.5 / 10
A meaningful difference from common solutions.

What the judges wrote

Claude Opus 5.5Preliminary61/100
Strengths
Your problem framing is concrete. The 180-day notice period buried in an email-delivery clause, set against a 60-day company rule, is exactly the kind of trap your target companies miss. Your README describes five working functions with worked examples: rule comparison, the law-change portfolio scan, cited Q&A with a 'not addressed' answer, notice-date reminders, and clause indexing. Your word-for-word quote verification with confirmed-quote counts and confidence levels is a real, code-level answer to hallucination risk in legal text. You are also notably honest about what is proven and what is not.
To improve
Your quality-testing evidence is mostly about engineering limits (rate limits, caching, key fallback) plus single synthetic examples with conflicts you seeded yourselves. Build the scored test set you mention now: dozens of contracts, measured precision and recall of findings, and quote-verification failure rates. Time one review done by hand and the same review with Signflow, so you can name the adoption metric the track asks for. Put running costs per contract and your data and language plans in the materials judges see, not only in BUSINESS.md.
GPT-6 AstraPreliminary63/100
Strengths
Your focus on hidden notice periods and changing legal obligations addresses concrete risks for companies with limited legal capacity. Your documented workflow combines clause citations, verbatim quote checks, abstention and explicit human review, making findings easier to audit. You also identify a specific engineering failure—free-tier limits—and describe relevant-clause retrieval and caching as fixes.
To improve
Strengthen your prototype evidence with a captured end-to-end run showing source clauses, detected conflicts and verified outputs, especially for the cross-referenced renewal example. Expand your seeded examples into a lawyer-labelled benchmark reporting missed conflicts, false positives, incorrect interpretations and abstention failures. In your pilot, compare review time and accuracy against manual review, and measure latency and running cost per contract and portfolio scan. Before using confidential contracts, define access controls, retention policies and how governing jurisdiction and legal applicability are checked.

The team

Hümmət Əzimzadə
developer
Tofiq Cahangirov
product
Toğrul Baxışov
data
Aslanlı Sadiq
designer
Mahmud ƏhmədovLead
developer

Disclosure

Required by the rules. Judges verify it against the repository.

Models: Gemini Flash, GPT-OSS-20b, Claude Opus 5.5
Data: Synthetic: all contracts, company terms and law texts are AI-generated for the demo, with deliberately seeded conflicts and no real or confidential data.
Components: We built it with Next.js, React, TypeScript and Tailwind CSS, with a Neon Postgres database, Google Gemini and Groq for the AI, and Vercel for hosting.
Built during the event: Confirmed

Quality testing

We kept hitting free-tier AI limits because every request sent all contracts to the model. We fixed it by splitting contracts into clauses in code, sending only the relevant clauses per question, caching every result so repeats cost nothing, and falling back automatically to other API keys when one runs out.

Worth 20 of the 100 points.

Running it

Setup intructions - Open the app and click Open dashboard. First go to Our terms, add or edit your company's rules, and save them. Then go to Contracts, click Add contract, upload a PDF or paste the text, choose the type, and click Extract clauses. To check a partner's contract, open Compare, pick the contract, and click Compare, then open each finding to see your rule, their clause, and suggested wording. To check a new law, open Law scanner, paste the law text, and click Scan to see which contracts are affected and how to fix them. To get an answer about a contract, open Ask and type your question in plain language, and the answer will show the clause it came from or tell you the contract does not cover it. To track deadlines, open Reminders, where red means act now, amber means coming up, and green means on track. A green tick under a quote means it was found word for word in the contract, and every result should be reviewed by a person before you act on it.

Project ownership

The team keeps the rights to this project. Judges were given access for verification only.