Internal platform
NGUERAIntelligencePlatform
Our internal operating platform — CRM, prospecting, quoting, invoicing, contracts and an AI assistant layer with mandatory human approval before sensitive actions. The platform has been engineered and technically verified; real-world usage results are still being measured.
Context
NGUERA TECH runs its own client work — projects, quotes, invoices, contracts — through a purpose-built internal platform rather than off-the-shelf tools. It was built to be the same kind of system we design for clients: role-based, audited, and structured around explicit human approval rather than automation for its own sake.
Internal problem
Managing prospecting, commercial preparation and delivery across a UK ↔ Senegal operation needed a single system — not a patchwork of spreadsheets and generic SaaS — with permissions that actually hold, and an audit trail for every sensitive action, technical or commercial.
Solution designed
A Next.js/Supabase platform combining a full operations portal (CRM, projects, quoting, invoicing, contracts, files, calendar) with an AI assistant layer built around one non-negotiable rule: an agent prepares, a human approves, only then does anything happen. Every tool call is checked server-side against role permissions, independently of what the model decides — the model is never trusted by default.
Modules available today
CRM
Clients
Projects
Quotes
Invoices
Contracts
Files
Calendar
Notifications
Settings
Security
Companies House prospecting
Explainable scoring
Commercial knowledge base
Dossier / email / call-script / quote-draft preparation
Commercial assistant tools
Architecture & technologies
Next.js 15 (App Router) · React 19 · TypeScript · Prisma 5 · Supabase (Postgres, Auth, Storage) · role-based access (Client / Team / Admin), re-verified server-side on every request · structured, schema-validated AI output before anything is written to the database.
Security & AI governance
Every AI tool call is logged. Cost and rate limits are enforced per user, in the database, not in memory. Every AI-generated draft — email, call script, quote, dossier — is prepared for human review; none is ever sent or executed automatically. Prompt-injection safeguards apply to any content an agent reads.
Current status
Core platform (CRM through commercial preparation tooling) is built, typed, and passes our standard verification (type-checking, linting, automated tests, production build). It has not yet processed a real prospect end-to-end with a live AI provider connected.
Next validations
Connect a live AI provider and run a first real batch through human review · validate the full prospect-to-dossier journey with every integration active · publish real content into the knowledge base before any AI-drafted output is used commercially.
Results
Results are still being measured — no usage or performance figures are published yet.