ScopeGuard AI
The User: Our primary users are Digital Agencies, Freelance Developers, Software Consultancies, and Project Managers who handle client-facing projects based on signed contracts. The Problem: We are solving a massive, multi-billion dollar industry problem known as "Scope Creep". During software development or design projects, clients constantly request "one small feature," "a quick button," or "minor changes" via Slack or emails. For agencies, this creates a toxic dilemma: They either say "No," which hurts the client relationship. Or they say "Yes" for free, resulting in thousands of hours of unpaid, unbilled work that destroys the agency's profit margins. Currently, manually checking a 50-page legal Statement of Work (SOW) PDF to verify if a feature is included takes too much time. Because of this friction, Project Managers often just accept the work for free. There is no automated, friction-free way to instantly check a request against a contract, estimate the out-of-scope cost, and generate a professional billable invoice to push back on the client diplomatically.
Preliminary score
- Value for the user19.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI20.5 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing10.0 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility7.5 / 15
Data requirements, running costs and a clear next step.
- Originality6.0 / 10
A meaningful difference from common solutions.
What the judges wrote
- Claude Opus 5.5Preliminary58/100
- Strengths
- Your focus on scope creep for agencies and freelancers is a real, recognisable pain, and you've turned it into a clear workflow: upload the SOW, audit the request, verify the evidence, estimate the cost, and export a Change Order. Using fuzzy matching to check that the quotes Gemini cites actually appear in the contract is a thoughtful anti-hallucination step that makes the verdict auditable. The four-way classification (in-scope, out-of-scope, partial, unaddressed) reflects how contracts really behave. A deployed Streamlit app plus a client-ready PDF output gives you a complete end-to-end scenario.
- To improve
- Your testing log is mostly UI and engineering bugs. Build a labelled set of real requests checked against 2–3 SOWs and report classification accuracy, including cases where the AI got the scope verdict wrong. Time a PM doing the same check manually versus with ScopeGuard, so you can replace the illustrative $500 slide with a measured outcome. Your Business Model slide is empty and running costs aren't stated, so add per-audit API cost, a pricing idea, and a pilot plan with one real agency. Adding an automatic retry with schema validation would fix the known malformed-JSON failure.
- GPT-6 AstraPreliminary68/100
- Strengths
- Your contract-to-change-order workflow addresses a concrete agency pain while keeping effort estimates and ambiguous scope decisions with the project manager. The four-way classification and quote-verification step distinguish the workflow from a generic contract chatbot. Your testing report names concrete widget-state failures and clearly acknowledges scanned-PDF and malformed-JSON limitations.
- To improve
- Show one complete SOW-to-change-order example and benchmark a labeled request set against manual review, reporting classification errors, review time, and quote-verification failures. Your fuzzy matching checks textual similarity, not whether a clause supports the scope decision, so add misleading-quote and ambiguous-clause tests. Before an agency pilot, document per-request running costs, contract retention and deletion, and user isolation for the disk cache, with authenticated access as a concrete next step.
The team
Amil RahimovLead
developer
Riad Memishov
developer
Ali Alizada
product
Mirhashim Bashirli
developer