Growenta AI
First-time entrepreneurs in Azerbaijan who are opening a small business — a cosmetics shop, a café, a clothing brand, an IT or education service — with a starting budget of roughly 5,000–100,000 AZN and no finance or marketing background. Today they need five things before they can start: a business plan, a realistic budget, a location, a brand, and a way to track money. Each comes from a different place. A consultant is slow and costs money they don't have. A spreadsheet does the maths but gives no advice. A general chatbot writes a convincing plan, but it knows nothing about their business, invents market figures with no source, and gets the arithmetic wrong. Nothing connects the plan to what happens after opening, so nobody notices when real income falls behind the forecast. The result: people start with a budget nobody checked, in a location nobody compared, and find out the plan was wrong only when the money runs out.
Preliminary score
- Value for the user19.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI24.0 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing14.5 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility12.0 / 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 5.5Preliminary74/100
- Strengths
- Your target user is specific: first-time Azerbaijani founders with 5,000–100,000 AZN and no finance background. The pains you name are concrete: unchecked budgets, unsourced chatbot figures and no link between the plan and actual results. Your architecture separates what the model does (writing and judging) from what code does (every sum, break-even and coordinate). It backs this with Zod validation, citation-id filtering, 'estimate' labels and a confirm-before-save transaction flow. Your testing write-up is honest and concrete: 21 unit tests including deliberately broken model output, real bugs found and fixed (nameless competitors, locale hydration mismatch, silent delete failure), and openly stated unfixed issues.
- To improve
- Your track asks for an adoption metric, but no outcome has been measured yet. Plan and analysis quality was checked by hand on one demo business, and the readiness score runs on placeholder market data. A small evaluation would make the core claim evidence rather than design: score 5–10 real or realistic plans, compare against a consultant or incubator assessment, and count invented or miscalculated figures versus a plain chatbot. Your cost section gives token sizes but no AZN or USD per user per month. Add that, plus a concrete source and schedule for loading real stat.gov.az data.
- GPT-6 AstraPreliminary77/100
- Strengths
- Your shared onboarding context and plan-versus-actual dashboard connect business planning to ongoing bookkeeping for Azerbaijani founders. Your architecture gives AI meaningful tasks while keeping financial calculations in code and requiring confirmation before saving transactions. Your documented tests and specific failure reports show thoughtful engineering, with unusually clear disclosure of placeholder data and unvalidated outputs.
- To improve
- Extend your deterministic tests with live-model evaluations across several businesses and both languages. Measure an adoption outcome, such as time to a usable, budget-checked plan, against the current spreadsheet/chatbot workflow. Replace DEMO market figures with verified, dated sources and validate readiness scores; checking citation IDs alone does not establish that a source supports a claim. Fix partial-month KPI comparisons and measure actual per-workflow costs, including images, search and retries.
The team
Nicat Bayramlı
business
Miraziz SeyidovLead
developer
Turgay Mammadov
data