Biznevo Documentation

How Biznevo works under the hood. The AI tools we orchestrate, the architecture we built, and how it all fits together.

Overview

Biznevo is an AI Employee that operates business operations end-to-end. Built on a custom orchestration engine that routes tasks across the world's most powerful AI models, Biznevo delivers professional results in marketing, lead generation, customer communication, social media, analytics, and continuous optimization.

This documentation covers the architecture, the AI tools we use, and how Biznevo achieves quality, security, and continuous improvement.

The Biznevo Engine

Our core orchestration layer.

The Biznevo Engine is a custom multi-stage pipeline that turns any customer request into a high-quality output. Every job flows through:

  1. 01

    Router

    Analyzes the request and routes it to the optimal AI model based on task type, complexity, and required capabilities.

  2. 02

    Generator

    Drafts the output using the assigned model with industry-specific prompts and customer context.

  3. 03

    Research Layer

    Pulls real-time market intelligence via Perplexity, ensuring every output is informed by current data.

  4. 04

    Critic

    Scores the output across 6 quality dimensions: specificity, accuracy, brand fit, completeness, professional tone, and actionability.

  5. 05

    Loop

    If the score does not meet our quality bar, the engine refines and retries until it does.

AI Tools We Orchestrate

Biznevo does not compete with the world's best AI tools. We orchestrate them.

Each tool below brings specialized capabilities. Biznevo's engine selects the right combination for every job.

Claude (Anthropic)

Our primary reasoning and writing model. Claude handles strategy, copywriting, complex analysis, and any task requiring nuanced understanding. We use Claude Opus for highest-tier reasoning and Claude Sonnet for high-volume tasks.

Perplexity

Real-time research and market intelligence. Perplexity's Sonar models pull fresh data on industries, competitors, audiences, and platform trends. Every job is informed by what is actually working in your market this week.

OpenAI (GPT models)

Specialized reasoning and image generation. We use OpenAI's models for tasks where their specific capabilities outperform alternatives, and gpt-image-1 to generate on-brand images for the websites and deliverables Biznevo builds, until Higgsfield integration completes.

Replicate

A cost-effective image-generation alternate. Biznevo routes image jobs to Replicate's hosted open models (such as FLUX) when they are the better fit on cost or style, so the visual engine never depends on a single provider.

Higgsfield

Coming soon

Production-grade image and video generation. Rather than relying on a single engine, Higgsfield aggregates over 30 state-of-the-art AI models: Google Veo 3.1, OpenAI Sora 2, Kling 3.0, Seedance 2.0, Nano Banana Pro, Cream 4.0, and more, into one workspace. Key capabilities Biznevo will use:

  • Cinema Studio: granular control over camera angles, lighting, and lens settings
  • SOUL ID: consistent character identity across dozens of generated clips
  • Motion & Expression Control: precise body and face animation
  • Sora 2 Trends: pre-tuned templates for current viral pacing on TikTok, Instagram, YouTube
  • Lipsync Studio: perfect audio-to-video synchronization
  • Outfit Swap, Upscaling, ElevenLabs Audio integration

Hermes

Coming soon

Biznevo's execution layer, coming after publishing ships: a master capability-learning system studies tools and workflows, every learned capability is approved by a human before use, and isolated per-customer sessions then operate your connected platforms strictly through official APIs and OAuth, with your tap required on every consequential action.

ElevenLabs

Coming soon

Voice AI for customer-facing calls. After MVP launches, Biznevo will handle inbound customer calls, qualify leads via voice, and respond to phone inquiries, all using ElevenLabs' voice synthesis.

Gemini's Video Model + Claude

Together, these analyze top-performing videos in your industry weekly, extracting hook formulas, retention techniques, and visual patterns. Biznevo then incorporates these into your video content automatically.

Security & Architecture

Biznevo is built for businesses that handle sensitive data. We follow enterprise security practices from day one.

  • Infrastructure: Hosted on Vercel for the application layer and AWS for storage (S3 for files), running on Neon Postgres for database. All infrastructure is enterprise-grade with redundancy and uptime SLAs.
  • Encryption: All data encrypted in transit (TLS 1.3) and at rest. Sensitive credentials encrypted with AES-256.
  • Multi-tenant safety: Postgres Row-Level Security enforced on every customer table. Customer A's data is mathematically inaccessible from Customer B's code path.
  • Sandboxed execution: Customer code, file uploads, and external integrations run in isolated environments. No customer can access another customer's resources.
  • OAuth-only: We never collect or store customer passwords for third-party platforms. All integrations use OAuth flows or official APIs.
  • No data selling: Customer data is never sold, never used to train models for other businesses, never shared with third parties for marketing purposes.
  • Compliance track: SOC-2 readiness, GDPR compliance, CCPA compliance. Audit logs for all admin actions.

Multi-Tenant Safety

How your business data stays isolated.

Biznevo serves multiple customers from the same infrastructure, which is standard for SaaS. What is not standard is how aggressively we enforce isolation. Every database query is filtered by your tenant_id at the Postgres layer, not in the application code where it could be bypassed. This means even if an application bug attempted to read another customer's data, the database itself refuses the request.

Quality Control

Every output Biznevo produces passes through our critic system. The Critic scores output on 6 dimensions:

  • Specificity: Is this specific to the customer's business, or generic?
  • Accuracy: Are the facts, claims, and references correct?
  • Brand fit: Does this match the customer's voice and style?
  • Completeness: Is anything missing that the customer asked for?
  • Professional tone: Does this read like work from a competent professional?
  • Actionability: Can the customer actually use this immediately?

Outputs that do not meet our threshold are refined and re-scored until they do. Quality is non-negotiable.

Per-Customer Learning

Biznevo learns your business with every job. Style Memory tracks your brand voice, audience preferences, and what works in your industry. After 20 jobs, Biznevo's output for your business is dramatically better than the first, and continuously improves from there. The learning is yours: your Style Memory is never used for other customers.

Continuous Improvement via Capability Radar

Biznevo monitors the AI ecosystem daily. When Anthropic, Perplexity, OpenAI, Gemini, Higgsfield, or any major AI provider releases new capabilities, our Capability Radar evaluates and integrates them automatically. This means your Biznevo gets more powerful every week, without you needing to learn or migrate to anything new.

Biznevo, your AI Business Employee