Architecture Review: Vect AI

Vect AI claims to be an “Execution-first marketing OS” that turns AI plans into workflows. Unlike standard chat interfaces that reset context after every session, Vect introduces a “State-Aware” architecture designed to maintain a persistent “Business Kernel” (brand voice, audience data, product truths) across all generated assets.

🛠️ The Tech Stack

Vect AI operates as an Agentic Orchestration Layer rather than a single model wrapper. It aggregates multiple specialized APIs into a cohesive workflow engine.

• Sensor Layer (Data Ingestion): Uses Google Search APIs (and likely Bing/Perplexity equivalents) for real-time “Market Signal Analysis.” This allows the system to ground its content in live web data rather than stale training sets.
• Strategist Layer (Logic): The core planning engine appears to be built on high-reasoning LLMs (likely GPT-4o or Claude 3.5 Sonnet) prompted to act as a “Campaign Builder.” It breaks high-level goals into structured 3-phase plans (Tease, Launch, Sustain).
• Creative Layer (Generation):
  • Video: Explicitly integrates Google Veo (and potentially Runway/Pika backups) for physics-aware, commercial-grade video generation (1080p, 60fps).
  • Images: Uses Google Imagen or Midjourney via API for asset creation.
• State Management (The “OS”): The differentiator is the “Business Kernel”-likely a Postgres/Supabase vector store that holds the user’s “Brand DNA.” This persistent state is injected into the context window of every agent, eliminating the need for repetitive prompting.

💰 Pricing Model

Vect AI currently operates on a Credit-Based / Waitlist model, reflecting the high compute costs of its underlying “Deep Research” and video generation features.

• Free/Trial: Limited access to the “Market Signal Analyzer” is often used as a lead magnet to demonstrate the “Blue Ocean” topic discovery capabilities.
• Credit System: Instead of a flat SaaS subscription, users likely purchase credits that are consumed based on the complexity of the task (e.g., generating a full video ad costs significantly more credits than drafting a blog post).
• Status: The tool is in active development (“Early Access”), with a focus on onboarding founders and technical marketers who need an “all-in-one” command center.

⚖️ Architect’s Verdict

Is this a Wrapper? Yes, but a “Thick Wrapper” (or Agentic Application). Vect AI does not train its own foundational models. However, dismissing it as “just a wrapper” ignores the complexity of its Orchestration Layer. By solving the “Context Fragmentation” problem (where you have to copy-paste context between ChatGPT, Midjourney, and specialized tools), it provides significant utility. It acts more like a specialized OS that manages context state across multiple ephemeral AI agents.

Developer Use Case: For developers, Vect AI is less of a “building block” and more of a “force multiplier” for their own product launches.

• No Public API (Yet): The focus is on the UI/UX for founders.
• Workflow Automation: It replaces the need for developers to hack together Python scripts using LangChain to automate their own marketing.
• Production Viability: Best suited for Solo Founders and Small Teams who need to execute marketing campaigns without hiring a dedicated agency. Enterprise teams might find the lack of granular control over the underlying prompts restrictive.