Innovation consulting · AI mobile app · Case study
NeuralX Inc.: AI business consultation app with a structured innovation method
React Native app and GraphQL backend on AWS guiding founders through a seven-dimension consultation, 45 innovation drivers and AI brainstorming.
- Client
- NeuralX Inc. is a technology company focused on innovation consulting and AI-driven business solutions; Treasure Island AI is its guided idea-development platform.
- Team
- Innovation Insight team with full-stack responsibilities
- Timeline
- 6 months from prototype to store release
- Services
- Mobile app development, AI and data, Cloud and DevOps, Web development
- 7
- consultation dimensions in the TIC method
- 45
- paradigm shift drivers across 7 categories
- 30
- AI brainstorming questions per idea
- 20
- supported industries
In short
NeuralX Inc. wanted its Treasure Island Consultation method turned into an AI-guided mobile product for founders. Innovation Insight built the React Native app, the GraphQL backend on AWS DynamoDB and Cognito, and the OpenAI-powered counsellor, brainstorming engine and report generator, with voice input, text-to-speech, multi-language support and full Mixpanel and Prometheus instrumentation.
Where they started.
- Entrepreneurs needed structured help to develop and validate business ideas, not an open-ended chatbot.
- The client's proprietary Treasure Island Consultation method has seven business dimensions and 45 paradigm shift drivers that had to be encoded as a guided workflow.
- AI conversations had to stay in character as an empathetic counsellor, keep context across a session and support voice in and out on mobile.
- Every idea, step, chat and report needed durable storage with strong consistency and full analytics on user behaviour.
The approach.
- Built a React Native app for iOS and Android with a chat interface, speech-to-text input, text-to-speech responses, charts and multi-language support.
- Implemented the TIC method as a five-phase workflow: industry selection, paradigm shift drivers, seven-dimension consultation, competitive benchmarking and AI brainstorming.
- Integrated OpenAI GPT-3.5-turbo as the 'Alice' counsellor persona with a twelve-message context window, tuned temperature and custom stop sequences.
- Designed an Express and Apollo Server GraphQL API with eight controller modules over AWS DynamoDB, Cognito authentication and Redis-backed subscriptions.
- Added a brainstorming engine that generates up to 30 contextual questions per idea and evaluates them across five categories, feeding a generated business plan report.
- Instrumented the stack with Mixpanel analytics, Prometheus metrics and CloudWatch logging, plus a Python Streamlit module for ML prototyping.
What shipped.
- React Native mobile app for iOS and Android with 15+ screens
- GraphQL API with 10+ queries and 11+ mutations
- AWS infrastructure: Cognito, DynamoDB (7 tables), CloudWatch
- OpenAI integration for chat consultation, question generation and recommendations
- TIC methodology and 45-driver paradigm shift framework implemented as data and workflow
- Report generation compiling consultation, benchmarking and brainstorming outputs
- Mixpanel analytics on frontend and backend, Prometheus and CloudWatch monitoring
- i18n multi-language support, voice input and text-to-speech
How the engagement ran.
Context
NeuralX Inc. runs innovation consulting built on the Treasure Island Consultation method: a structured path through business overview, vision, market size, target customers, value proposition, unique selling point and business model, informed by 45 paradigm shift drivers such as AI, circular economy and subscription models. The company wanted that expertise available to any founder through a mobile app with an AI counsellor. Innovation Insight delivered the product with full-stack responsibility for architecture, backend, mobile, AI integration and cloud infrastructure.
What we built
- A guided idea workflow: choose one of 20 industries, select relevant drivers from 45 options in seven categories, complete the seven-dimension consultation, benchmark against three to five companies with AI recommendations, then answer up to 30 generated brainstorming questions scored across market opportunity, USP, value proposition, sustainability and execution feasibility.
- An AI chat counsellor, Alice, on GPT-3.5-turbo with a twelve-message context, persistent DynamoDB storage with sequential numbering, and voice input and text-to-speech on mobile.
- Report generation that compiles consultation data, benchmarking and brainstorming into a business plan available through the API.
- Authentication on AWS Cognito with sign-up, login, automatic token refresh and password reset, plus profile settings, language selection and voice preferences.
- Analytics with Mixpanel on both client and server: session tracking, user identification, idea creation, step completion and feature engagement.
Architecture and stack
| Layer | Technology | Why |
|---|---|---|
| Mobile | React Native 0.78, React 19, TypeScript, Apollo Client, React Navigation, Gifted Chat, Gifted Charts, Reanimated, MMKV, i18next | Cross-platform app with chat, charts, animation, fast local storage and multi-language support |
| API | Express, Apollo Server, GraphQL, eight controller modules | Single typed API for ideas, steps, chat, drivers, industries and reports |
| Data | AWS DynamoDB with seven tables and strong-consistency reads | Idea-centric access patterns with composite keys for steps and ordered chat |
| AI | OpenAI GPT-3.5-turbo and GPT-4 via API, temperature 0.7, bounded context | Consultation chat, question generation and recommendations |
| Auth and real time | AWS Cognito, Redis subscriptions, WebSocket | Managed identity and live GraphQL updates |
| Observability | Prometheus, log4js with CloudWatch, Mixpanel | Latency and error metrics, centralised logs and product analytics |
How we worked
The method came first: each dimension has a status flow from pending to confirmed, each phase has its own data, and every table in DynamoDB maps to one of those concepts. With the data model settled, the API, mobile screens and AI prompts were built against it in parallel. The mobile app is organised by feature, with authentication, dashboard, chat, ideas, settings and onboarding modules, and the backend follows a controller pattern with a separate AI library and analytics library.
A Python Streamlit module gave the client a place to prototype analysis workflows quickly before they were productised in the API, which kept experimentation cheap without touching the mobile release cycle.
Scale and complexity
| Metric | Value |
|---|---|
| Mobile screens | 15+ |
| GraphQL operations | 10+ queries, 11+ mutations |
| DynamoDB tables | 7 |
| Consultation dimensions and phases | 7 dimensions, 5 phases |
| Paradigm shift drivers | 45 across 7 categories |
| Industries | 20 |
| Brainstorming questions per idea | Up to 30, evaluated in 5 categories |
What this means for your project
Expertise-driven products, whether consulting, coaching, training or advisory, can be turned into software that scales beyond the founder's calendar. The key is to encode the method precisely and use AI for the conversational parts, with guardrails, persistence and analytics so the experience improves over time.
Innovation Insight builds these AI-guided products end to end, mobile app and backend included, and can prototype the AI conversation on your method within the first two to three weeks.
What changed.
- Production-grade AI mobile platform delivered with cloud-native architecture, real-time subscriptions and full analytics
- Proprietary consultation method fully digitised across seven dimensions, 45 drivers and 20 industries.
- Up to 30 AI-generated brainstorming questions per idea, evaluated across five categories.
- Live GraphQL endpoint at api.tigame.ai.
Questions about this project.
Can you turn a proprietary consulting or coaching method into an app?
Yes. The work is to model the method as data and states, then guide the user through it with AI where judgement is needed. Treasure Island AI encodes seven dimensions, five phases and 45 drivers, each with its own status flow and prompts.
How do you keep an AI assistant on topic and in character?
With a system prompt that defines the persona and boundaries, a bounded context window (twelve messages here), a conservative temperature, custom stop sequences and persistent storage of every exchange so the conversation can be audited and improved.
Why DynamoDB and Cognito rather than a relational database?
The data model is idea-centric with predictable access patterns, so DynamoDB's composite keys and strong-consistency reads fit well and scale without operations work. Cognito handles sign-up, tokens and password resets without custom auth code.